Skip to content

GitLab

Commit e9cee049 authored by luopl's avatar luopl
Browse files

Initial commit

parents
Pipeline #1056 canceled with stages
Hide whitespace changes
Inline Side-by-side
Showing with 2592 additions and 0 deletions
configs/pretrain_v1/yolo_world_l_dual_vlpan_l2norm_2e-3_100e_4x8gpus_obj365v1_goldg_train_lvis_val.py 0 → 100644
View file @ e9cee049
_base_ = ('../../third_party/mmyolo/configs/yolov8/'
'yolov8_l_syncbn_fast_8xb16-500e_coco.py')
custom_imports = dict(imports=['yolo_world'],
allow_failed_imports=False)
# hyper-parameters
num_classes = 1203
num_training_classes = 80
max_epochs = 100 # Maximum training epochs
close_mosaic_epochs = 2
save_epoch_intervals = 2
text_channels = 512
neck_embed_channels = [128, 256, _base_.last_stage_out_channels // 2]
neck_num_heads = [4, 8, _base_.last_stage_out_channels // 2 // 32]
base_lr = 2e-3
weight_decay = 0.05 / 2
train_batch_size_per_gpu = 16
# model settings
model = dict(
type='YOLOWorldDetector',
mm_neck=True,
num_train_classes=num_training_classes,
num_test_classes=num_classes,
data_preprocessor=dict(type='YOLOWDetDataPreprocessor'),
backbone=dict(
_delete_=True,
type='MultiModalYOLOBackbone',
image_model={{_base_.model.backbone}},
text_model=dict(
type='HuggingCLIPLanguageBackbone',
model_name='openai/clip-vit-base-patch32',
frozen_modules=['all'])),
neck=dict(type='YOLOWorldDualPAFPN',
guide_channels=text_channels,
embed_channels=neck_embed_channels,
num_heads=neck_num_heads,
block_cfg=dict(type='MaxSigmoidCSPLayerWithTwoConv'),
text_enhancder=dict(type='ImagePoolingAttentionModule',
embed_channels=256,
num_heads=8)),
bbox_head=dict(type='YOLOWorldHead',
head_module=dict(type='YOLOWorldHeadModule',
embed_dims=text_channels,
num_classes=num_training_classes)),
train_cfg=dict(assigner=dict(num_classes=num_training_classes)))
# dataset settings
text_transform = [
dict(type='RandomLoadText',
num_neg_samples=(num_classes, num_classes),
max_num_samples=num_training_classes,
padding_to_max=True,
padding_value=''),
dict(type='mmdet.PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape', 'flip',
'flip_direction', 'texts'))
]
train_pipeline = [
*_base_.pre_transform,
dict(type='MultiModalMosaic',
img_scale=_base_.img_scale,
pad_val=114.0,
pre_transform=_base_.pre_transform),
dict(
type='YOLOv5RandomAffine',
max_rotate_degree=0.0,
max_shear_degree=0.0,
scaling_ratio_range=(1 - _base_.affine_scale, 1 + _base_.affine_scale),
max_aspect_ratio=_base_.max_aspect_ratio,
border=(-_base_.img_scale[0] // 2, -_base_.img_scale[1] // 2),
border_val=(114, 114, 114)),
*_base_.last_transform[:-1],
*text_transform,
]
train_pipeline_stage2 = [*_base_.train_pipeline_stage2[:-1], *text_transform]
obj365v1_train_dataset = dict(
type='MultiModalDataset',
dataset=dict(
type='YOLOv5Objects365V1Dataset',
data_root='data/objects365v1/',
ann_file='annotations/objects365_train.json',
data_prefix=dict(img='train/'),
filter_cfg=dict(filter_empty_gt=False, min_size=32)),
class_text_path='data/texts/obj365v1_class_texts.json',
pipeline=train_pipeline)
mg_train_dataset = dict(type='YOLOv5MixedGroundingDataset',
data_root='data/mixed_grounding/',
ann_file='annotations/final_mixed_train_no_coco.json',
data_prefix=dict(img='gqa/images/'),
filter_cfg=dict(filter_empty_gt=False, min_size=32),
pipeline=train_pipeline)
flickr_train_dataset = dict(
type='YOLOv5MixedGroundingDataset',
data_root='data/flickr/',
ann_file='annotations/final_flickr_separateGT_train.json',
data_prefix=dict(img='full_images/'),
filter_cfg=dict(filter_empty_gt=True, min_size=32),
pipeline=train_pipeline)
train_dataloader = dict(batch_size=train_batch_size_per_gpu,
collate_fn=dict(type='yolow_collate'),
dataset=dict(_delete_=True,
type='ConcatDataset',
datasets=[
obj365v1_train_dataset,
flickr_train_dataset, mg_train_dataset
],
ignore_keys=['classes', 'palette']))
test_pipeline = [
*_base_.test_pipeline[:-1],
dict(type='LoadText'),
dict(type='mmdet.PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape',
'scale_factor', 'pad_param', 'texts'))
]
coco_val_dataset = dict(
_delete_=True,
type='MultiModalDataset',
dataset=dict(type='YOLOv5LVISV1Dataset',
data_root='data/coco/',
test_mode=True,
ann_file='lvis/lvis_v1_val.json',
data_prefix=dict(img=''),
batch_shapes_cfg=None),
class_text_path='data/texts/lvis_v1_class_texts.json',
pipeline=test_pipeline)
val_dataloader = dict(dataset=coco_val_dataset)
test_dataloader = val_dataloader
val_evaluator = dict(type='mmdet.LVISMetric',
ann_file='data/coco/lvis/lvis_v1_val.json',
metric='bbox')
test_evaluator = val_evaluator
# training settings
default_hooks = dict(param_scheduler=dict(max_epochs=max_epochs),
checkpoint=dict(interval=save_epoch_intervals,
rule='greater'))
custom_hooks = [
dict(type='EMAHook',
ema_type='ExpMomentumEMA',
momentum=0.0001,
update_buffers=True,
strict_load=False,
priority=49),
dict(type='mmdet.PipelineSwitchHook',
switch_epoch=max_epochs - close_mosaic_epochs,
switch_pipeline=train_pipeline_stage2)
]
train_cfg = dict(max_epochs=max_epochs,
val_interval=10,
dynamic_intervals=[((max_epochs - close_mosaic_epochs),
_base_.val_interval_stage2)])
optim_wrapper = dict(optimizer=dict(
_delete_=True,
type='AdamW',
lr=base_lr,
weight_decay=weight_decay,
batch_size_per_gpu=train_batch_size_per_gpu),
paramwise_cfg=dict(bias_decay_mult=0.0,
norm_decay_mult=0.0,
custom_keys={
'backbone.text_model':
dict(lr_mult=0.01),
'logit_scale':
dict(weight_decay=0.0)
}),
constructor='YOLOWv5OptimizerConstructor')
configs/pretrain_v1/yolo_world_m_dual_vlpan_l2norm_2e-3_100e_4x8gpus_obj365v1_goldg_train_lvis_minival.py 0 → 100644
View file @ e9cee049
_base_ = ('../../third_party/mmyolo/configs/yolov8/'
'yolov8_m_syncbn_fast_8xb16-500e_coco.py')
custom_imports = dict(imports=['yolo_world'],
allow_failed_imports=False)
# hyper-parameters
num_classes = 1203
num_training_classes = 80
max_epochs = 100 # Maximum training epochs
close_mosaic_epochs = 2
save_epoch_intervals = 2
text_channels = 512
neck_embed_channels = [128, 256, _base_.last_stage_out_channels // 2]
neck_num_heads = [4, 8, _base_.last_stage_out_channels // 2 // 32]
base_lr = 2e-3
weight_decay = 0.05 / 2
train_batch_size_per_gpu = 16
# model settings
model = dict(
type='YOLOWorldDetector',
mm_neck=True,
num_train_classes=num_training_classes,
num_test_classes=num_classes,
data_preprocessor=dict(type='YOLOWDetDataPreprocessor'),
backbone=dict(
_delete_=True,
type='MultiModalYOLOBackbone',
image_model={{_base_.model.backbone}},
text_model=dict(
type='HuggingCLIPLanguageBackbone',
model_name='openai/clip-vit-base-patch32',
frozen_modules=['all'])),
neck=dict(type='YOLOWorldDualPAFPN',
guide_channels=text_channels,
embed_channels=neck_embed_channels,
num_heads=neck_num_heads,
block_cfg=dict(type='MaxSigmoidCSPLayerWithTwoConv'),
text_enhancder=dict(type='ImagePoolingAttentionModule',
embed_channels=256,
num_heads=8)),
bbox_head=dict(type='YOLOWorldHead',
head_module=dict(type='YOLOWorldHeadModule',
embed_dims=text_channels,
num_classes=num_training_classes)),
train_cfg=dict(assigner=dict(num_classes=num_training_classes)))
# dataset settings
text_transform = [
dict(type='RandomLoadText',
num_neg_samples=(num_classes, num_classes),
max_num_samples=num_training_classes,
padding_to_max=True,
padding_value=''),
dict(type='mmdet.PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape', 'flip',
'flip_direction', 'texts'))
]
train_pipeline = [
*_base_.pre_transform,
dict(type='MultiModalMosaic',
img_scale=_base_.img_scale,
pad_val=114.0,
pre_transform=_base_.pre_transform),
dict(
type='YOLOv5RandomAffine',
max_rotate_degree=0.0,
max_shear_degree=0.0,
scaling_ratio_range=(1 - _base_.affine_scale, 1 + _base_.affine_scale),
max_aspect_ratio=_base_.max_aspect_ratio,
border=(-_base_.img_scale[0] // 2, -_base_.img_scale[1] // 2),
border_val=(114, 114, 114)),
*_base_.last_transform[:-1],
*text_transform,
]
train_pipeline_stage2 = [*_base_.train_pipeline_stage2[:-1], *text_transform]
obj365v1_train_dataset = dict(
type='MultiModalDataset',
dataset=dict(
type='YOLOv5Objects365V1Dataset',
data_root='data/objects365v1/',
ann_file='annotations/objects365_train.json',
data_prefix=dict(img='train/'),
filter_cfg=dict(filter_empty_gt=False, min_size=32)),
class_text_path='data/texts/obj365v1_class_texts.json',
pipeline=train_pipeline)
mg_train_dataset = dict(type='YOLOv5MixedGroundingDataset',
data_root='data/mixed_grounding/',
ann_file='annotations/final_mixed_train_no_coco.json',
data_prefix=dict(img='gqa/images/'),
filter_cfg=dict(filter_empty_gt=False, min_size=32),
pipeline=train_pipeline)
flickr_train_dataset = dict(
type='YOLOv5MixedGroundingDataset',
data_root='data/flickr/',
ann_file='annotations/final_flickr_separateGT_train.json',
data_prefix=dict(img='full_images/'),
filter_cfg=dict(filter_empty_gt=True, min_size=32),
pipeline=train_pipeline)
train_dataloader = dict(batch_size=train_batch_size_per_gpu,
collate_fn=dict(type='yolow_collate'),
dataset=dict(_delete_=True,
type='ConcatDataset',
datasets=[
obj365v1_train_dataset,
flickr_train_dataset, mg_train_dataset
],
ignore_keys=['classes', 'palette']))
test_pipeline = [
*_base_.test_pipeline[:-1],
dict(type='LoadText'),
dict(type='mmdet.PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape',
'scale_factor', 'pad_param', 'texts'))
]
coco_val_dataset = dict(
_delete_=True,
type='MultiModalDataset',
dataset=dict(type='YOLOv5LVISV1Dataset',
data_root='data/coco/',
test_mode=True,
ann_file='lvis/lvis_v1_minival_inserted_image_name.json',
data_prefix=dict(img=''),
batch_shapes_cfg=None),
class_text_path='data/texts/lvis_v1_class_texts.json',
pipeline=test_pipeline)
val_dataloader = dict(dataset=coco_val_dataset)
test_dataloader = val_dataloader
val_evaluator = dict(type='mmdet.LVISMetric',
ann_file='data/coco/lvis/lvis_v1_minival_inserted_image_name.json',
metric='bbox')
test_evaluator = val_evaluator
# training settings
default_hooks = dict(param_scheduler=dict(max_epochs=max_epochs),
checkpoint=dict(interval=save_epoch_intervals,
rule='greater'))
custom_hooks = [
dict(type='EMAHook',
ema_type='ExpMomentumEMA',
momentum=0.0001,
update_buffers=True,
strict_load=False,
priority=49),
dict(type='mmdet.PipelineSwitchHook',
switch_epoch=max_epochs - close_mosaic_epochs,
switch_pipeline=train_pipeline_stage2)
]
train_cfg = dict(max_epochs=max_epochs,
val_interval=10,
dynamic_intervals=[((max_epochs - close_mosaic_epochs),
_base_.val_interval_stage2)])
optim_wrapper = dict(optimizer=dict(
_delete_=True,
type='AdamW',
lr=base_lr,
weight_decay=weight_decay,
batch_size_per_gpu=train_batch_size_per_gpu),
paramwise_cfg=dict(bias_decay_mult=0.0,
norm_decay_mult=0.0,
custom_keys={
'backbone.text_model':
dict(lr_mult=0.01),
'logit_scale':
dict(weight_decay=0.0)
}),
constructor='YOLOWv5OptimizerConstructor')
configs/pretrain_v1/yolo_world_s_dual_vlpan_l2norm_2e-3_100e_4x8gpus_obj365v1_goldg_train_lvis_minival.py 0 → 100644
View file @ e9cee049
_base_ = ('../../third_party/mmyolo/configs/yolov8/'
'yolov8_s_syncbn_fast_8xb16-500e_coco.py')
custom_imports = dict(imports=['yolo_world'],
allow_failed_imports=False)
# hyper-parameters
num_classes = 1203
num_training_classes = 80
max_epochs = 100 # Maximum training epochs
close_mosaic_epochs = 2
save_epoch_intervals = 2
text_channels = 512
neck_embed_channels = [128, 256, _base_.last_stage_out_channels // 2]
neck_num_heads = [4, 8, _base_.last_stage_out_channels // 2 // 32]
base_lr = 2e-3
weight_decay = 0.05 / 2
train_batch_size_per_gpu = 16
# model settings
model = dict(
type='YOLOWorldDetector',
mm_neck=True,
num_train_classes=num_training_classes,
num_test_classes=num_classes,
data_preprocessor=dict(type='YOLOWDetDataPreprocessor'),
backbone=dict(
_delete_=True,
type='MultiModalYOLOBackbone',
image_model={{_base_.model.backbone}},
text_model=dict(
type='HuggingCLIPLanguageBackbone',
model_name='openai/clip-vit-base-patch32',
frozen_modules=['all'])),
neck=dict(type='YOLOWorldDualPAFPN',
guide_channels=text_channels,
embed_channels=neck_embed_channels,
num_heads=neck_num_heads,
block_cfg=dict(type='MaxSigmoidCSPLayerWithTwoConv'),
text_enhancder=dict(type='ImagePoolingAttentionModule',
embed_channels=256,
num_heads=8)),
bbox_head=dict(type='YOLOWorldHead',
head_module=dict(type='YOLOWorldHeadModule',
embed_dims=text_channels,
num_classes=num_training_classes)),
train_cfg=dict(assigner=dict(num_classes=num_training_classes)))
# dataset settings
text_transform = [
dict(type='RandomLoadText',
num_neg_samples=(num_classes, num_classes),
max_num_samples=num_training_classes,
padding_to_max=True,
padding_value=''),
dict(type='mmdet.PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape', 'flip',
'flip_direction', 'texts'))
]
train_pipeline = [
*_base_.pre_transform,
dict(type='MultiModalMosaic',
img_scale=_base_.img_scale,
pad_val=114.0,
pre_transform=_base_.pre_transform),
dict(
type='YOLOv5RandomAffine',
max_rotate_degree=0.0,
max_shear_degree=0.0,
scaling_ratio_range=(1 - _base_.affine_scale, 1 + _base_.affine_scale),
max_aspect_ratio=_base_.max_aspect_ratio,
border=(-_base_.img_scale[0] // 2, -_base_.img_scale[1] // 2),
border_val=(114, 114, 114)),
*_base_.last_transform[:-1],
*text_transform,
]
train_pipeline_stage2 = [*_base_.train_pipeline_stage2[:-1], *text_transform]
obj365v1_train_dataset = dict(
type='MultiModalDataset',
dataset=dict(
type='YOLOv5Objects365V1Dataset',
data_root='data/objects365v1/',
ann_file='annotations/objects365_train.json',
data_prefix=dict(img='train/'),
filter_cfg=dict(filter_empty_gt=False, min_size=32)),
class_text_path='data/texts/obj365v1_class_texts.json',
pipeline=train_pipeline)
mg_train_dataset = dict(type='YOLOv5MixedGroundingDataset',
data_root='data/mixed_grounding/',
ann_file='annotations/final_mixed_train_no_coco.json',
data_prefix=dict(img='gqa/images/'),
filter_cfg=dict(filter_empty_gt=False, min_size=32),
pipeline=train_pipeline)
flickr_train_dataset = dict(
type='YOLOv5MixedGroundingDataset',
data_root='data/flickr/',
ann_file='annotations/final_flickr_separateGT_train.json',
data_prefix=dict(img='full_images/'),
filter_cfg=dict(filter_empty_gt=True, min_size=32),
pipeline=train_pipeline)
train_dataloader = dict(batch_size=train_batch_size_per_gpu,
collate_fn=dict(type='yolow_collate'),
dataset=dict(_delete_=True,
type='ConcatDataset',
datasets=[
obj365v1_train_dataset,
flickr_train_dataset, mg_train_dataset
],
ignore_keys=['classes', 'palette']))
test_pipeline = [
*_base_.test_pipeline[:-1],
dict(type='LoadText'),
dict(type='mmdet.PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape',
'scale_factor', 'pad_param', 'texts'))
]
coco_val_dataset = dict(
_delete_=True,
type='MultiModalDataset',
dataset=dict(type='YOLOv5LVISV1Dataset',
data_root='data/coco/',
test_mode=True,
ann_file='lvis/lvis_v1_minival_inserted_image_name.json',
data_prefix=dict(img=''),
batch_shapes_cfg=None),
class_text_path='data/texts/lvis_v1_class_texts.json',
pipeline=test_pipeline)
val_dataloader = dict(dataset=coco_val_dataset)
test_dataloader = val_dataloader
val_evaluator = dict(type='mmdet.LVISMetric',
ann_file='data/coco/lvis/lvis_v1_minival_inserted_image_name.json',
metric='bbox')
test_evaluator = val_evaluator
# training settings
default_hooks = dict(param_scheduler=dict(max_epochs=max_epochs),
checkpoint=dict(interval=save_epoch_intervals,
rule='greater'))
custom_hooks = [
dict(type='EMAHook',
ema_type='ExpMomentumEMA',
momentum=0.0001,
update_buffers=True,
strict_load=False,
priority=49),
dict(type='mmdet.PipelineSwitchHook',
switch_epoch=max_epochs - close_mosaic_epochs,
switch_pipeline=train_pipeline_stage2)
]
train_cfg = dict(max_epochs=max_epochs,
val_interval=10,
dynamic_intervals=[((max_epochs - close_mosaic_epochs),
_base_.val_interval_stage2)])
optim_wrapper = dict(optimizer=dict(
_delete_=True,
type='AdamW',
lr=base_lr,
weight_decay=weight_decay,
batch_size_per_gpu=train_batch_size_per_gpu),
paramwise_cfg=dict(bias_decay_mult=0.0,
norm_decay_mult=0.0,
custom_keys={
'backbone.text_model':
dict(lr_mult=0.01),
'logit_scale':
dict(weight_decay=0.0)
}),
constructor='YOLOWv5OptimizerConstructor')
configs/pretrain_v1/yolo_world_x_dual_vlpan_l2norm_2e-3_100e_4x8gpus_obj365v1_goldg_train_lvis_minival.py 0 → 100644
View file @ e9cee049
_base_ = ('../../third_party/mmyolo/configs/yolov8/'
'yolov8_x_syncbn_fast_8xb16-500e_coco.py')
custom_imports = dict(imports=['yolo_world'],
allow_failed_imports=False)
# hyper-parameters
num_classes = 1203
num_training_classes = 80
max_epochs = 100 # Maximum training epochs
close_mosaic_epochs = 2
save_epoch_intervals = 2
text_channels = 512
neck_embed_channels = [128, 256, _base_.last_stage_out_channels // 2]
neck_num_heads = [4, 8, _base_.last_stage_out_channels // 2 // 32]
base_lr = 2e-3
weight_decay = 0.05 / 2
train_batch_size_per_gpu = 16
# model settings
model = dict(
type='YOLOWorldDetector',
mm_neck=True,
num_train_classes=num_training_classes,
num_test_classes=num_classes,
data_preprocessor=dict(type='YOLOWDetDataPreprocessor'),
backbone=dict(
_delete_=True,
type='MultiModalYOLOBackbone',
image_model={{_base_.model.backbone}},
text_model=dict(
type='HuggingCLIPLanguageBackbone',
model_name='openai/clip-vit-base-patch32',
frozen_modules=['all'])),
neck=dict(type='YOLOWorldDualPAFPN',
guide_channels=text_channels,
embed_channels=neck_embed_channels,
num_heads=neck_num_heads,
block_cfg=dict(type='MaxSigmoidCSPLayerWithTwoConv'),
text_enhancder=dict(type='ImagePoolingAttentionModule',
embed_channels=256,
num_heads=8)),
bbox_head=dict(type='YOLOWorldHead',
head_module=dict(type='YOLOWorldHeadModule',
embed_dims=text_channels,
num_classes=num_training_classes)),
train_cfg=dict(assigner=dict(num_classes=num_training_classes)))
# dataset settings
text_transform = [
dict(type='RandomLoadText',
num_neg_samples=(num_classes, num_classes),
max_num_samples=num_training_classes,
padding_to_max=True,
padding_value=''),
dict(type='mmdet.PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape', 'flip',
'flip_direction', 'texts'))
]
train_pipeline = [
*_base_.pre_transform,
dict(type='MultiModalMosaic',
img_scale=_base_.img_scale,
pad_val=114.0,
pre_transform=_base_.pre_transform),
dict(
type='YOLOv5RandomAffine',
max_rotate_degree=0.0,
max_shear_degree=0.0,
scaling_ratio_range=(1 - _base_.affine_scale, 1 + _base_.affine_scale),
max_aspect_ratio=_base_.max_aspect_ratio,
border=(-_base_.img_scale[0] // 2, -_base_.img_scale[1] // 2),
border_val=(114, 114, 114)),
*_base_.last_transform[:-1],
*text_transform,
]
train_pipeline_stage2 = [*_base_.train_pipeline_stage2[:-1], *text_transform]
obj365v1_train_dataset = dict(
type='MultiModalDataset',
dataset=dict(
type='YOLOv5Objects365V1Dataset',
data_root='data/objects365v1/',
ann_file='annotations/objects365_train.json',
data_prefix=dict(img='train/'),
filter_cfg=dict(filter_empty_gt=False, min_size=32)),
class_text_path='data/texts/obj365v1_class_texts.json',
pipeline=train_pipeline)
mg_train_dataset = dict(type='YOLOv5MixedGroundingDataset',
data_root='data/mixed_grounding/',
ann_file='annotations/final_mixed_train_no_coco.json',
data_prefix=dict(img='gqa/images/'),
filter_cfg=dict(filter_empty_gt=False, min_size=32),
pipeline=train_pipeline)
flickr_train_dataset = dict(
type='YOLOv5MixedGroundingDataset',
data_root='data/flickr/',
ann_file='annotations/final_flickr_separateGT_train.json',
data_prefix=dict(img='full_images/'),
filter_cfg=dict(filter_empty_gt=True, min_size=32),
pipeline=train_pipeline)
train_dataloader = dict(batch_size=train_batch_size_per_gpu,
collate_fn=dict(type='yolow_collate'),
dataset=dict(_delete_=True,
type='ConcatDataset',
datasets=[
obj365v1_train_dataset,
flickr_train_dataset, mg_train_dataset
],
ignore_keys=['classes', 'palette']))
test_pipeline = [
*_base_.test_pipeline[:-1],
dict(type='LoadText'),
dict(type='mmdet.PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape',
'scale_factor', 'pad_param', 'texts'))
]
coco_val_dataset = dict(
_delete_=True,
type='MultiModalDataset',
dataset=dict(type='YOLOv5LVISV1Dataset',
data_root='data/coco/',
test_mode=True,
ann_file='lvis/lvis_v1_minival_inserted_image_name.json',
data_prefix=dict(img=''),
batch_shapes_cfg=None),
class_text_path='data/texts/lvis_v1_class_texts.json',
pipeline=test_pipeline)
val_dataloader = dict(dataset=coco_val_dataset)
test_dataloader = val_dataloader
val_evaluator = dict(type='mmdet.LVISMetric',
ann_file='data/coco/lvis/lvis_v1_minival_inserted_image_name.json',
metric='bbox')
test_evaluator = val_evaluator
# training settings
default_hooks = dict(param_scheduler=dict(max_epochs=max_epochs),
checkpoint=dict(interval=save_epoch_intervals,
rule='greater'))
custom_hooks = [
dict(type='EMAHook',
ema_type='ExpMomentumEMA',
momentum=0.0001,
update_buffers=True,
strict_load=False,
priority=49),
dict(type='mmdet.PipelineSwitchHook',
switch_epoch=max_epochs - close_mosaic_epochs,
switch_pipeline=train_pipeline_stage2)
]
train_cfg = dict(max_epochs=max_epochs,
val_interval=10,
dynamic_intervals=[((max_epochs - close_mosaic_epochs),
_base_.val_interval_stage2)])
optim_wrapper = dict(optimizer=dict(
_delete_=True,
type='AdamW',
lr=base_lr,
weight_decay=weight_decay,
batch_size_per_gpu=train_batch_size_per_gpu),
paramwise_cfg=dict(bias_decay_mult=0.0,
norm_decay_mult=0.0,
custom_keys={
'backbone.text_model':
dict(lr_mult=0.01),
'logit_scale':
dict(weight_decay=0.0)
}),
constructor='YOLOWv5OptimizerConstructor')
configs/prompt_tuning_coco/READEME.md 0 → 100644
View file @ e9cee049
## Prompt Tuning for YOLO-World
### NOTE:
This folder contains many experimental config files, which will be removed later!!
### Experimental Results
| Model | Config | AP | AP50 | AP75 | APS | APM | APL |
| :---- | :----: | :--: | :--: | :---: | :-: | :-: | :-: |
| YOLO-World-v2-L | Zero-shot | 45.7 | 61.6 | 49.8 | 29.9 | 50.0 | 60.8 |
| [YOLO-World-v2-L](./../configs/prompt_tuning_coco/yolo_world_v2_l_vlpan_bn_2e-4_80e_8gpus_mask-refine_prompt_tuning_coco.py) | Prompt tuning | 47.9 | 64.3 | 52.5 | 31.9 | 52.6 | 61.3 |
configs/prompt_tuning_coco/yolo_world_v2_l_vlpan_bn_2e-4_80e_8gpus_mask-refine_prompt_tuning_coco.py 0 → 100644
View file @ e9cee049
_base_ = ('../../third_party/mmyolo/configs/yolov8/'
'yolov8_l_mask-refine_syncbn_fast_8xb16-500e_coco.py')
custom_imports = dict(imports=['yolo_world'], allow_failed_imports=False)
# hyper-parameters
num_classes = 80
num_training_classes = 80
max_epochs = 80 # Maximum training epochs
close_mosaic_epochs = 10
save_epoch_intervals = 5
text_channels = 512
neck_embed_channels = [128, 256, _base_.last_stage_out_channels // 2]
neck_num_heads = [4, 8, _base_.last_stage_out_channels // 2 // 32]
base_lr = 2e-3
weight_decay = 0.05
train_batch_size_per_gpu = 16
load_from = 'pretrained_models/yolo_world_l_clip_t2i_bn_2e-3adamw_32xb16-100e_obj365v1_goldg_cc3mlite_train-ca93cd1f.pth'
persistent_workers = False
# model settings
model = dict(type='SimpleYOLOWorldDetector',
mm_neck=True,
num_train_classes=num_training_classes,
num_test_classes=num_classes,
embedding_path='embeddings/clip_vit_b32_coco_80_embeddings.npy',
prompt_dim=text_channels,
num_prompts=80,
data_preprocessor=dict(type='YOLOv5DetDataPreprocessor'),
backbone=dict(_delete_=True,
type='MultiModalYOLOBackbone',
text_model=None,
image_model={{_base_.model.backbone}},
frozen_stages=4,
with_text_model=False),
neck=dict(type='YOLOWorldPAFPN',
freeze_all=True,
guide_channels=text_channels,
embed_channels=neck_embed_channels,
num_heads=neck_num_heads,
block_cfg=dict(type='MaxSigmoidCSPLayerWithTwoConv')),
bbox_head=dict(type='YOLOWorldHead',
head_module=dict(
type='YOLOWorldHeadModule',
freeze_all=True,
use_bn_head=True,
embed_dims=text_channels,
num_classes=num_training_classes)),
train_cfg=dict(assigner=dict(num_classes=num_training_classes)))
# dataset settings
final_transform = [
dict(type='mmdet.PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape', 'flip',
'flip_direction'))
]
mosaic_affine_transform = [
dict(type='Mosaic',
img_scale=_base_.img_scale,
pad_val=114.0,
pre_transform=_base_.pre_transform),
dict(type='YOLOv5CopyPaste', prob=_base_.copypaste_prob),
dict(
type='YOLOv5RandomAffine',
max_rotate_degree=0.0,
max_shear_degree=0.0,
max_aspect_ratio=100.,
scaling_ratio_range=(1 - _base_.affine_scale, 1 + _base_.affine_scale),
# img_scale is (width, height)
border=(-_base_.img_scale[0] // 2, -_base_.img_scale[1] // 2),
border_val=(114, 114, 114),
min_area_ratio=_base_.min_area_ratio,
use_mask_refine=_base_.use_mask2refine)
]
train_pipeline = [
*_base_.pre_transform, *mosaic_affine_transform,
dict(type='YOLOv5MixUp',
prob=_base_.mixup_prob,
pre_transform=[*_base_.pre_transform, *mosaic_affine_transform]),
*_base_.last_transform[:-1], *final_transform
]
train_pipeline_stage2 = [*_base_.train_pipeline_stage2[:-1], *final_transform]
coco_train_dataset = dict(type='YOLOv5CocoDataset',
data_root='data/coco',
ann_file='annotations/instances_train2017.json',
data_prefix=dict(img='train2017/'),
filter_cfg=dict(filter_empty_gt=False, min_size=32),
pipeline=train_pipeline)
train_dataloader = dict(persistent_workers=persistent_workers,
batch_size=train_batch_size_per_gpu,
collate_fn=dict(type='yolow_collate'),
dataset=coco_train_dataset)
train_dataloader = dict(persistent_workers=persistent_workers,
batch_size=train_batch_size_per_gpu,
collate_fn=dict(type='yolow_collate'),
dataset=coco_train_dataset)
test_pipeline = [
*_base_.test_pipeline[:-1],
dict(type='mmdet.PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape',
'scale_factor', 'pad_param'))
]
coco_val_dataset = dict(type='YOLOv5CocoDataset',
data_root='data/coco',
ann_file='annotations/instances_val2017.json',
data_prefix=dict(img='val2017/'),
filter_cfg=dict(filter_empty_gt=False, min_size=32),
pipeline=test_pipeline)
val_dataloader = dict(dataset=coco_val_dataset)
test_dataloader = val_dataloader
# training settings
default_hooks = dict(param_scheduler=dict(scheduler_type='linear',
lr_factor=0.01,
max_epochs=max_epochs),
checkpoint=dict(max_keep_ckpts=-1,
save_best=None,
interval=save_epoch_intervals))
custom_hooks = [
dict(type='EMAHook',
ema_type='ExpMomentumEMA',
momentum=0.0001,
update_buffers=True,
strict_load=False,
priority=49),
dict(type='mmdet.PipelineSwitchHook',
switch_epoch=max_epochs - close_mosaic_epochs,
switch_pipeline=train_pipeline_stage2)
]
train_cfg = dict(max_epochs=max_epochs,
val_interval=5,
dynamic_intervals=[((max_epochs - close_mosaic_epochs),
_base_.val_interval_stage2)])
optim_wrapper = dict(optimizer=dict(
_delete_=True,
type='AdamW',
lr=base_lr,
weight_decay=weight_decay,
batch_size_per_gpu=train_batch_size_per_gpu),
paramwise_cfg=dict(bias_decay_mult=0.0,
norm_decay_mult=0.0,
custom_keys={
'backbone.text_model':
dict(lr_mult=0.01),
'logit_scale':
dict(weight_decay=0.0),
'embeddings':
dict(weight_decay=0.0)
}),
constructor='YOLOWv5OptimizerConstructor')
# evaluation settings
val_evaluator = dict(_delete_=True,
type='mmdet.CocoMetric',
proposal_nums=(100, 1, 10),
ann_file='data/coco/annotations/instances_val2017.json',
metric='bbox')
find_unused_parameters = True
configs/prompt_tuning_coco/yolo_world_v2_l_vlpan_bn_2e-4_80e_8gpus_prompt_tuning_coco.py 0 → 100644
View file @ e9cee049
_base_ = ('../../third_party/mmyolo/configs/yolov8/'
'yolov8_l_syncbn_fast_8xb16-500e_coco.py')
custom_imports = dict(imports=['yolo_world'], allow_failed_imports=False)
# hyper-parameters
num_classes = 80
num_training_classes = 80
max_epochs = 80 # Maximum training epochs
close_mosaic_epochs = 10
save_epoch_intervals = 5
text_channels = 512
neck_embed_channels = [128, 256, _base_.last_stage_out_channels // 2]
neck_num_heads = [4, 8, _base_.last_stage_out_channels // 2 // 32]
base_lr = 2e-4
weight_decay = 0.05
train_batch_size_per_gpu = 16
load_from = 'pretrained_models/yolo_world_l_clip_t2i_bn_2e-3adamw_32xb16-100e_obj365v1_goldg_cc3mlite_train-ca93cd1f.pth'
persistent_workers = False
# model settings
model = dict(type='SimpleYOLOWorldDetector',
mm_neck=True,
num_train_classes=num_training_classes,
num_test_classes=num_classes,
embedding_path='embeddings/clip_vit_b32_coco_80_embeddings.npy',
prompt_dim=text_channels,
num_prompts=80,
freeze_prompt=False,
data_preprocessor=dict(type='YOLOv5DetDataPreprocessor'),
backbone=dict(_delete_=True,
type='MultiModalYOLOBackbone',
text_model=None,
image_model={{_base_.model.backbone}},
frozen_stages=4,
with_text_model=False),
neck=dict(type='YOLOWorldPAFPN',
freeze_all=True,
guide_channels=text_channels,
embed_channels=neck_embed_channels,
num_heads=neck_num_heads,
block_cfg=dict(type='MaxSigmoidCSPLayerWithTwoConv')),
bbox_head=dict(type='YOLOWorldHead',
head_module=dict(
type='YOLOWorldHeadModule',
freeze_all=True,
use_bn_head=True,
embed_dims=text_channels,
num_classes=num_training_classes)),
train_cfg=dict(assigner=dict(num_classes=num_training_classes)))
# dataset settings
coco_train_dataset = dict(type='YOLOv5CocoDataset',
data_root='data/coco',
ann_file='annotations/instances_train2017.json',
data_prefix=dict(img='train2017/'),
filter_cfg=dict(filter_empty_gt=False, min_size=32),
pipeline=_base_.train_pipeline)
train_dataloader = dict(persistent_workers=persistent_workers,
batch_size=train_batch_size_per_gpu,
collate_fn=dict(type='yolow_collate'),
dataset=coco_train_dataset)
coco_val_dataset = dict(type='YOLOv5CocoDataset',
data_root='data/coco',
ann_file='annotations/instances_val2017.json',
data_prefix=dict(img='val2017/'),
filter_cfg=dict(filter_empty_gt=False, min_size=32),
pipeline=_base_.test_pipeline)
val_dataloader = dict(dataset=coco_val_dataset)
test_dataloader = val_dataloader
# training settings
default_hooks = dict(param_scheduler=dict(scheduler_type='linear',
lr_factor=0.01,
max_epochs=max_epochs),
checkpoint=dict(max_keep_ckpts=-1,
save_best=None,
interval=save_epoch_intervals))
custom_hooks = [
dict(type='EMAHook',
ema_type='ExpMomentumEMA',
momentum=0.0001,
update_buffers=True,
strict_load=False,
priority=49),
dict(type='mmdet.PipelineSwitchHook',
switch_epoch=max_epochs - close_mosaic_epochs,
switch_pipeline=_base_.train_pipeline_stage2)
]
train_cfg = dict(max_epochs=max_epochs,
val_interval=5,
dynamic_intervals=[((max_epochs - close_mosaic_epochs),
_base_.val_interval_stage2)])
optim_wrapper = dict(optimizer=dict(
_delete_=True,
type='AdamW',
lr=base_lr,
weight_decay=weight_decay,
batch_size_per_gpu=train_batch_size_per_gpu),
paramwise_cfg=dict(custom_keys={
'backbone.text_model':
dict(lr_mult=0.01),
'logit_scale':
dict(weight_decay=0.0),
'embeddings':
dict(weight_decay=0.0)
}),
constructor='YOLOWv5OptimizerConstructor')
# evaluation settings
val_evaluator = dict(_delete_=True,
type='mmdet.CocoMetric',
proposal_nums=(100, 1, 10),
ann_file='data/coco/annotations/instances_val2017.json',
metric='bbox')
configs/prompt_tuning_coco/yolo_world_v2_l_vlpan_bn_sgd_1e-3_80e_8gpus_all_finetuning_coco.py 0 → 100644
View file @ e9cee049
_base_ = ('../../third_party/mmyolo/configs/yolov8/'
'yolov8_l_syncbn_fast_8xb16-500e_coco.py')
custom_imports = dict(imports=['yolo_world'], allow_failed_imports=False)
# hyper-parameters
num_classes = 80
num_training_classes = 80
max_epochs = 40 # Maximum training epochs
close_mosaic_epochs = 10
save_epoch_intervals = 5
text_channels = 512
neck_embed_channels = [128, 256, _base_.last_stage_out_channels // 2]
neck_num_heads = [4, 8, _base_.last_stage_out_channels // 2 // 32]
base_lr = 1e-3
weight_decay = 0.0005
train_batch_size_per_gpu = 16
load_from = 'pretrained_models/yolo_world_l_clip_t2i_bn_2e-3adamw_32xb16-100e_obj365v1_goldg_cc3mlite_train-ca93cd1f.pth'
persistent_workers = False
# model settings
model = dict(type='SimpleYOLOWorldDetector',
mm_neck=True,
num_train_classes=num_training_classes,
num_test_classes=num_classes,
embedding_path='embeddings/clip_vit_b32_coco_80_embeddings.npy',
prompt_dim=text_channels,
num_prompts=80,
freeze_prompt=True,
data_preprocessor=dict(type='YOLOv5DetDataPreprocessor'),
backbone=dict(_delete_=True,
type='MultiModalYOLOBackbone',
text_model=None,
image_model={{_base_.model.backbone}},
with_text_model=False),
neck=dict(type='YOLOWorldPAFPN',
freeze_all=False,
guide_channels=text_channels,
embed_channels=neck_embed_channels,
num_heads=neck_num_heads,
block_cfg=dict(type='MaxSigmoidCSPLayerWithTwoConv')),
bbox_head=dict(type='YOLOWorldHead',
head_module=dict(
type='YOLOWorldHeadModule',
freeze_all=False,
use_bn_head=True,
embed_dims=text_channels,
num_classes=num_training_classes)),
train_cfg=dict(assigner=dict(num_classes=num_training_classes)))
# dataset settings
coco_train_dataset = dict(type='YOLOv5CocoDataset',
data_root='data/coco',
ann_file='annotations/instances_train2017.json',
data_prefix=dict(img='train2017/'),
filter_cfg=dict(filter_empty_gt=False, min_size=32),
pipeline=_base_.train_pipeline)
train_dataloader = dict(persistent_workers=persistent_workers,
batch_size=train_batch_size_per_gpu,
collate_fn=dict(type='yolow_collate'),
dataset=coco_train_dataset)
coco_val_dataset = dict(type='YOLOv5CocoDataset',
data_root='data/coco',
ann_file='annotations/instances_val2017.json',
data_prefix=dict(img='val2017/'),
filter_cfg=dict(filter_empty_gt=False, min_size=32),
pipeline=_base_.test_pipeline)
val_dataloader = dict(dataset=coco_val_dataset)
test_dataloader = val_dataloader
# training settings
default_hooks = dict(param_scheduler=dict(scheduler_type='linear',
lr_factor=0.01,
max_epochs=max_epochs),
checkpoint=dict(max_keep_ckpts=-1,
save_best=None,
interval=save_epoch_intervals))
custom_hooks = [
dict(type='EMAHook',
ema_type='ExpMomentumEMA',
momentum=0.0001,
update_buffers=True,
strict_load=False,
priority=49),
dict(type='mmdet.PipelineSwitchHook',
switch_epoch=max_epochs - close_mosaic_epochs,
switch_pipeline=_base_.train_pipeline_stage2)
]
train_cfg = dict(max_epochs=max_epochs,
val_interval=5,
dynamic_intervals=[((max_epochs - close_mosaic_epochs),
_base_.val_interval_stage2)])
optim_wrapper = dict(optimizer=dict(
_delete_=True,
type='SGD',
lr=base_lr,
momentum=0.937,
nesterov=True,
weight_decay=weight_decay,
batch_size_per_gpu=train_batch_size_per_gpu))
# evaluation settings
val_evaluator = dict(_delete_=True,
type='mmdet.CocoMetric',
proposal_nums=(100, 1, 10),
ann_file='data/coco/annotations/instances_val2017.json',
metric='bbox')
configs/segmentation/README.md 0 → 100644
View file @ e9cee049
## Fine-tuning YOLO-World for Instance Segmentation
### Models
We fine-tune YOLO-World on LVIS (`LVIS-Base`) with mask annotations for open-vocabulary (zero-shot) instance segmentation.
We provide two fine-tuning strategies YOLO-World towards open-vocabulary instance segmentation:
* fine-tuning `all modules`: leads to better LVIS segmentation accuracy but affects the zero-shot performance.
* fine-tuning the `segmentation head`: maintains the zero-shot performanc but lowers LVIS segmentation accuracy.
| Model | Fine-tuning Data | Fine-tuning Modules| AP<sup>mask</su> | AP<sub>r</sub> | AP<sub>c</sub> | AP<sub>f</sub> | Weights |
| :---- | :--------------- | :----------------: | :--------------: | :------------: | :------------: | :------------: | :-----: |
| [YOLO-World-Seg-M](./yolo_world_seg_m_dual_vlpan_2e-4_80e_8gpus_allmodules_finetune_lvis.py) | `LVIS-Base` | `all modules` | 25.9 | 13.4 | 24.9 | 32.6 | [HF Checkpoints 🤗](https://huggingface.co/wondervictor/YOLO-World/blob/main/yolo_world_seg_m_dual_vlpan_2e-4_80e_8gpus_allmodules_finetune_lvis-ca465825.pth) |
| [YOLO-World-v2-Seg-M](./yolo_world_seg_m_dual_vlpan_2e-4_80e_8gpus_allmodules_finetune_lvis.py) | `LVIS-Base` | `all modules` | 25.9 | 13.4 | 24.9 | 32.6 | [HF Checkpoints 🤗]() |
| [YOLO-World-Seg-L](./yolo_world_seg_l_dual_vlpan_2e-4_80e_8gpus_allmodules_finetune_lvis.py) | `LVIS-Base` | `all modules` | 28.7 | 15.0 | 28.3 | 35.2| [HF Checkpoints 🤗](https://huggingface.co/wondervictor/YOLO-World/blob/main/yolo_world_seg_l_dual_vlpan_2e-4_80e_8gpus_allmodules_finetune_lvis-8c58c916.pth) |
| [YOLO-World-v2-Seg-L](./yolo_world_seg_l_dual_vlpan_2e-4_80e_8gpus_allmodules_finetune_lvis.py) | `LVIS-Base` | `all modules` | 28.7 | 15.0 | 28.3 | 35.2| [HF Checkpoints 🤗]() |
| [YOLO-World-Seg-M](./yolo_seg_world_m_dual_vlpan_2e-4_80e_8gpus_seghead_finetune_lvis.py) | `LVIS-Base` | `seg head` | 16.7 | 12.6 | 14.6 | 20.8 | [HF Checkpoints 🤗](https://huggingface.co/wondervictor/YOLO-World/blob/main/yolo_world_seg_m_dual_vlpan_2e-4_80e_8gpus_seghead_finetune_lvis-7bca59a7.pth) |
| [YOLO-World-v2-Seg-M](./yolo_world_v2_seg_m_vlpan_bn_2e-4_80e_8gpus_seghead_finetune_lvis.py) | `LVIS-Base` | `seg head` | 17.8 | 13.9 | 15.5 | 22.0 | [HF Checkpoints 🤗]() |
| [YOLO-World-Seg-L](yolo_seg_world_l_dual_vlpan_2e-4_80e_8gpus_seghead_finetune_lvis.py) | `LVIS-Base` | `seg head` | 19.1 | 14.2 | 17.2 | 23.5 | [HF Checkpoints 🤗](https://huggingface.co/wondervictor/YOLO-World/blob/main/yolo_world_seg_l_dual_vlpan_2e-4_80e_8gpus_seghead_finetune_lvis-5a642d30.pth) |
| [YOLO-World-v2-Seg-L](./yolo_world_v2_seg_l_vlpan_bn_2e-4_80e_8gpus_seghead_finetune_lvis.py) | `LVIS-Base` | `seg head` | 19.8 | 17.2 | 17.5 | 23.6 | [HF Checkpoints 🤗]() |
**NOTE:**
1. The mask AP are evaluated on the LVIS `val 1.0`.
2. All models are fine-tuned for 80 epochs on `LVIS-Base` (866 categories, `common + frequent`).
3. The YOLO-World-Seg with only `seg head` fine-tuned maintains the original zero-shot detection capability and segments objects.
configs/segmentation/yolo_world_seg_l_dual_vlpan_2e-4_80e_8gpus_allmodules_finetune_lvis.py 0 → 100644
View file @ e9cee049
_base_ = (
'../../third_party/mmyolo/configs/yolov8/yolov8_l_mask-refine_syncbn_fast_8xb16-500e_coco.py'
)
custom_imports = dict(imports=['yolo_world'], allow_failed_imports=False)
# hyper-parameters
num_classes = 1203
num_training_classes = 80
max_epochs = 80 # Maximum training epochs
close_mosaic_epochs = 10
save_epoch_intervals = 5
text_channels = 512
neck_embed_channels = [128, 256, _base_.last_stage_out_channels // 2]
neck_num_heads = [4, 8, _base_.last_stage_out_channels // 2 // 32]
base_lr = 2e-4
weight_decay = 0.05
train_batch_size_per_gpu = 8
load_from = 'pretrained_models/yolo_world_l_clip_base_dual_vlpan_2e-3adamw_32xb16_100e_o365_goldg_train_pretrained-0e566235.pth'
persistent_workers = False
text_model_name = '../pretrained_models/clip-vit-base-patch32-projection'
# text_model_name = 'openai/clip-vit-base-patch32'
# Polygon2Mask
downsample_ratio = 4
mask_overlap = False
use_mask2refine = True
max_aspect_ratio = 100
min_area_ratio = 0.01
# model settings
model = dict(
type='YOLOWorldDetector',
mm_neck=True,
num_train_classes=num_training_classes,
num_test_classes=num_classes,
data_preprocessor=dict(type='YOLOWDetDataPreprocessor'),
backbone=dict(
_delete_=True,
type='MultiModalYOLOBackbone',
image_model={{_base_.model.backbone}},
text_model=dict(
type='HuggingCLIPLanguageBackbone',
model_name=text_model_name,
frozen_modules=[])),
neck=dict(type='YOLOWorldDualPAFPN',
guide_channels=text_channels,
embed_channels=neck_embed_channels,
num_heads=neck_num_heads,
block_cfg=dict(type='MaxSigmoidCSPLayerWithTwoConv'),
text_enhancder=dict(type='ImagePoolingAttentionModule',
embed_channels=256,
num_heads=8)),
bbox_head=dict(type='YOLOWorldSegHead',
head_module=dict(type='YOLOWorldSegHeadModule',
embed_dims=text_channels,
num_classes=num_training_classes,
mask_channels=32,
proto_channels=256),
mask_overlap=mask_overlap,
loss_mask=dict(type='mmdet.CrossEntropyLoss',
use_sigmoid=True,
reduction='none'),
loss_mask_weight=1.0),
train_cfg=dict(assigner=dict(num_classes=num_training_classes)),
test_cfg=dict(mask_thr_binary=0.5, fast_test=True))
pre_transform = [
dict(type='LoadImageFromFile', backend_args=_base_.backend_args),
dict(type='LoadAnnotations',
with_bbox=True,
with_mask=True,
mask2bbox=True)
]
last_transform = [
dict(type='mmdet.Albu',
transforms=_base_.albu_train_transforms,
bbox_params=dict(type='BboxParams',
format='pascal_voc',
label_fields=['gt_bboxes_labels',
'gt_ignore_flags']),
keymap={
'img': 'image',
'gt_bboxes': 'bboxes'
}),
dict(type='YOLOv5HSVRandomAug'),
dict(type='mmdet.RandomFlip', prob=0.5),
dict(type='Polygon2Mask',
downsample_ratio=downsample_ratio,
mask_overlap=mask_overlap),
]
# dataset settings
text_transform = [
dict(type='RandomLoadText',
num_neg_samples=(num_classes, num_classes),
max_num_samples=num_training_classes,
padding_to_max=True,
padding_value=''),
dict(type='PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape', 'flip',
'flip_direction', 'texts'))
]
mosaic_affine_transform = [
dict(type='MultiModalMosaic',
img_scale=_base_.img_scale,
pad_val=114.0,
pre_transform=pre_transform),
dict(type='YOLOv5CopyPaste', prob=_base_.copypaste_prob),
dict(
type='YOLOv5RandomAffine',
max_rotate_degree=0.0,
max_shear_degree=0.0,
max_aspect_ratio=100.,
scaling_ratio_range=(1 - _base_.affine_scale, 1 + _base_.affine_scale),
# img_scale is (width, height)
border=(-_base_.img_scale[0] // 2, -_base_.img_scale[1] // 2),
border_val=(114, 114, 114),
min_area_ratio=_base_.min_area_ratio,
use_mask_refine=True)
]
train_pipeline = [
*pre_transform, *mosaic_affine_transform,
dict(type='YOLOv5MultiModalMixUp',
prob=_base_.mixup_prob,
pre_transform=[*pre_transform, *mosaic_affine_transform]),
*last_transform, *text_transform
]
_train_pipeline_stage2 = [
*pre_transform,
dict(type='YOLOv5KeepRatioResize', scale=_base_.img_scale),
dict(type='LetterResize',
scale=_base_.img_scale,
allow_scale_up=True,
pad_val=dict(img=114.0)),
dict(type='YOLOv5RandomAffine',
max_rotate_degree=0.0,
max_shear_degree=0.0,
scaling_ratio_range=(1 - _base_.affine_scale,
1 + _base_.affine_scale),
max_aspect_ratio=_base_.max_aspect_ratio,
border_val=(114, 114, 114),
min_area_ratio=min_area_ratio,
use_mask_refine=use_mask2refine), *last_transform
]
train_pipeline_stage2 = [*_train_pipeline_stage2, *text_transform]
coco_train_dataset = dict(
_delete_=True,
type='MultiModalDataset',
dataset=dict(type='YOLOv5LVISV1Dataset',
data_root='data/coco',
ann_file='lvis/lvis_v1_train_base.json',
data_prefix=dict(img=''),
filter_cfg=dict(filter_empty_gt=True, min_size=32)),
class_text_path='data/texts/lvis_v1_base_class_texts.json',
pipeline=train_pipeline)
train_dataloader = dict(persistent_workers=persistent_workers,
batch_size=train_batch_size_per_gpu,
collate_fn=dict(type='yolow_collate'),
dataset=coco_train_dataset)
test_pipeline = [
*_base_.test_pipeline[:-1],
dict(type='LoadText'),
dict(type='mmdet.PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape',
'scale_factor', 'pad_param', 'texts'))
]
# training settings
default_hooks = dict(param_scheduler=dict(scheduler_type='linear',
lr_factor=0.01,
max_epochs=max_epochs),
checkpoint=dict(max_keep_ckpts=-1,
save_best=None,
interval=save_epoch_intervals))
custom_hooks = [
dict(type='EMAHook',
ema_type='ExpMomentumEMA',
momentum=0.0001,
update_buffers=True,
strict_load=False,
priority=49),
dict(type='mmdet.PipelineSwitchHook',
switch_epoch=max_epochs - close_mosaic_epochs,
switch_pipeline=train_pipeline_stage2)
]
train_cfg = dict(max_epochs=max_epochs,
val_interval=5,
dynamic_intervals=[((max_epochs - close_mosaic_epochs),
_base_.val_interval_stage2)])
optim_wrapper = dict(optimizer=dict(
_delete_=True,
type='AdamW',
lr=base_lr,
weight_decay=weight_decay,
batch_size_per_gpu=train_batch_size_per_gpu),
paramwise_cfg=dict(bias_decay_mult=0.0,
norm_decay_mult=0.0,
custom_keys={
'backbone.text_model':
dict(lr_mult=0.01),
'logit_scale':
dict(weight_decay=0.0),
}),
constructor='YOLOWv5OptimizerConstructor')
# evaluation settings
coco_val_dataset = dict(
_delete_=True,
type='MultiModalDataset',
dataset=dict(type='YOLOv5LVISV1Dataset',
data_root='data/coco/',
test_mode=True,
ann_file='lvis/lvis_v1_val.json',
data_prefix=dict(img=''),
batch_shapes_cfg=None),
class_text_path='data/captions/lvis_v1_class_captions.json',
pipeline=test_pipeline)
val_dataloader = dict(dataset=coco_val_dataset)
test_dataloader = val_dataloader
val_evaluator = dict(type='mmdet.LVISMetric',
ann_file='data/coco/lvis/lvis_v1_val.json',
metric=['bbox', 'segm'])
test_evaluator = val_evaluator
find_unused_parameters = True
configs/segmentation/yolo_world_seg_l_dual_vlpan_2e-4_80e_8gpus_seghead_finetune_lvis.py 0 → 100644
View file @ e9cee049
_base_ = (
'../../third_party/mmyolo/configs/yolov8/yolov8_l_mask-refine_syncbn_fast_8xb16-500e_coco.py'
)
custom_imports = dict(imports=['yolo_world'], allow_failed_imports=False)
# hyper-parameters
num_classes = 1203
num_training_classes = 80
max_epochs = 80 # Maximum training epochs
close_mosaic_epochs = 10
save_epoch_intervals = 5
text_channels = 512
neck_embed_channels = [128, 256, _base_.last_stage_out_channels // 2]
neck_num_heads = [4, 8, _base_.last_stage_out_channels // 2 // 32]
base_lr = 2e-4
weight_decay = 0.05
train_batch_size_per_gpu = 8
load_from = 'pretrained_models/yolo_world_l_clip_base_dual_vlpan_2e-3adamw_32xb16_100e_o365_goldg_train_pretrained-0e566235.pth'
persistent_workers = False
# Polygon2Mask
downsample_ratio = 4
mask_overlap = False
use_mask2refine = True
max_aspect_ratio = 100
min_area_ratio = 0.01
# model settings
model = dict(
type='YOLOWorldDetector',
mm_neck=True,
num_train_classes=num_training_classes,
num_test_classes=num_classes,
data_preprocessor=dict(type='YOLOWDetDataPreprocessor'),
backbone=dict(
_delete_=True,
type='MultiModalYOLOBackbone',
image_model={{_base_.model.backbone}},
frozen_stages=4, # frozen the image backbone
text_model=dict(
type='HuggingCLIPLanguageBackbone',
model_name='openai/clip-vit-base-patch32',
frozen_modules=['all'])),
neck=dict(type='YOLOWorldDualPAFPN',
freeze_all=True,
guide_channels=text_channels,
embed_channels=neck_embed_channels,
num_heads=neck_num_heads,
block_cfg=dict(type='MaxSigmoidCSPLayerWithTwoConv'),
text_enhancder=dict(type='ImagePoolingAttentionModule',
embed_channels=256,
num_heads=8)),
bbox_head=dict(type='YOLOWorldSegHead',
head_module=dict(type='YOLOWorldSegHeadModule',
embed_dims=text_channels,
num_classes=num_training_classes,
mask_channels=32,
proto_channels=256,
freeze_bbox=True),
mask_overlap=mask_overlap,
loss_mask=dict(type='mmdet.CrossEntropyLoss',
use_sigmoid=True,
reduction='none'),
loss_mask_weight=1.0),
train_cfg=dict(assigner=dict(num_classes=num_training_classes)),
test_cfg=dict(mask_thr_binary=0.5, fast_test=True))
pre_transform = [
dict(type='LoadImageFromFile', backend_args=_base_.backend_args),
dict(type='LoadAnnotations',
with_bbox=True,
with_mask=True,
mask2bbox=True)
]
last_transform = [
dict(type='mmdet.Albu',
transforms=_base_.albu_train_transforms,
bbox_params=dict(type='BboxParams',
format='pascal_voc',
label_fields=['gt_bboxes_labels',
'gt_ignore_flags']),
keymap={
'img': 'image',
'gt_bboxes': 'bboxes'
}),
dict(type='YOLOv5HSVRandomAug'),
dict(type='mmdet.RandomFlip', prob=0.5),
dict(type='Polygon2Mask',
downsample_ratio=downsample_ratio,
mask_overlap=mask_overlap),
]
# dataset settings
text_transform = [
dict(type='RandomLoadText',
num_neg_samples=(num_classes, num_classes),
max_num_samples=num_training_classes,
padding_to_max=True,
padding_value=''),
dict(type='PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape', 'flip',
'flip_direction', 'texts'))
]
mosaic_affine_transform = [
dict(type='MultiModalMosaic',
img_scale=_base_.img_scale,
pad_val=114.0,
pre_transform=pre_transform),
dict(type='YOLOv5CopyPaste', prob=_base_.copypaste_prob),
dict(
type='YOLOv5RandomAffine',
max_rotate_degree=0.0,
max_shear_degree=0.0,
max_aspect_ratio=100.,
scaling_ratio_range=(1 - _base_.affine_scale, 1 + _base_.affine_scale),
# img_scale is (width, height)
border=(-_base_.img_scale[0] // 2, -_base_.img_scale[1] // 2),
border_val=(114, 114, 114),
min_area_ratio=_base_.min_area_ratio,
use_mask_refine=True)
]
train_pipeline = [
*pre_transform, *mosaic_affine_transform,
dict(type='YOLOv5MultiModalMixUp',
prob=_base_.mixup_prob,
pre_transform=[*pre_transform, *mosaic_affine_transform]),
*last_transform, *text_transform
]
_train_pipeline_stage2 = [
*pre_transform,
dict(type='YOLOv5KeepRatioResize', scale=_base_.img_scale),
dict(type='LetterResize',
scale=_base_.img_scale,
allow_scale_up=True,
pad_val=dict(img=114.0)),
dict(type='YOLOv5RandomAffine',
max_rotate_degree=0.0,
max_shear_degree=0.0,
scaling_ratio_range=(1 - _base_.affine_scale,
1 + _base_.affine_scale),
max_aspect_ratio=_base_.max_aspect_ratio,
border_val=(114, 114, 114),
min_area_ratio=min_area_ratio,
use_mask_refine=use_mask2refine), *last_transform
]
train_pipeline_stage2 = [*_train_pipeline_stage2, *text_transform]
coco_train_dataset = dict(
_delete_=True,
type='MultiModalDataset',
dataset=dict(type='YOLOv5LVISV1Dataset',
data_root='data/coco',
ann_file='lvis/lvis_v1_train_base.json',
data_prefix=dict(img=''),
filter_cfg=dict(filter_empty_gt=True, min_size=32)),
class_text_path='data/texts/lvis_v1_base_class_texts.json',
pipeline=train_pipeline)
train_dataloader = dict(persistent_workers=persistent_workers,
batch_size=train_batch_size_per_gpu,
collate_fn=dict(type='yolow_collate'),
dataset=coco_train_dataset)
test_pipeline = [
*_base_.test_pipeline[:-1],
dict(type='LoadText'),
dict(type='mmdet.PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape',
'scale_factor', 'pad_param', 'texts'))
]
# training settings
default_hooks = dict(param_scheduler=dict(scheduler_type='linear',
lr_factor=0.01,
max_epochs=max_epochs),
checkpoint=dict(max_keep_ckpts=-1,
save_best=None,
interval=save_epoch_intervals))
custom_hooks = [
dict(type='EMAHook',
ema_type='ExpMomentumEMA',
momentum=0.0001,
update_buffers=True,
strict_load=False,
priority=49),
dict(type='mmdet.PipelineSwitchHook',
switch_epoch=max_epochs - close_mosaic_epochs,
switch_pipeline=train_pipeline_stage2)
]
train_cfg = dict(max_epochs=max_epochs,
val_interval=5,
dynamic_intervals=[((max_epochs - close_mosaic_epochs),
_base_.val_interval_stage2)])
optim_wrapper = dict(optimizer=dict(
_delete_=True,
type='AdamW',
lr=base_lr,
weight_decay=weight_decay,
batch_size_per_gpu=train_batch_size_per_gpu),
paramwise_cfg=dict(bias_decay_mult=0.0,
norm_decay_mult=0.0,
custom_keys={
'backbone.text_model':
dict(lr_mult=0.01),
'logit_scale':
dict(weight_decay=0.0),
'neck':
dict(lr_mult=0.0),
'head.head_module.reg_preds':
dict(lr_mult=0.0),
'head.head_module.cls_preds':
dict(lr_mult=0.0),
'head.head_module.cls_contrasts':
dict(lr_mult=0.0)
}),
constructor='YOLOWv5OptimizerConstructor')
# evaluation settings
coco_val_dataset = dict(
_delete_=True,
type='MultiModalDataset',
dataset=dict(type='YOLOv5LVISV1Dataset',
data_root='data/coco/',
test_mode=True,
ann_file='lvis/lvis_v1_val.json',
data_prefix=dict(img=''),
batch_shapes_cfg=None),
class_text_path='data/captions/lvis_v1_class_captions.json',
pipeline=test_pipeline)
val_dataloader = dict(dataset=coco_val_dataset)
test_dataloader = val_dataloader
val_evaluator = dict(type='mmdet.LVISMetric',
ann_file='data/coco/lvis/lvis_v1_val.json',
metric=['bbox', 'segm'])
test_evaluator = val_evaluator
find_unused_parameters = True
configs/segmentation/yolo_world_seg_m_dual_vlpan_2e-4_80e_8gpus_allmodules_finetune_lvis.py 0 → 100644
View file @ e9cee049
_base_ = (
'../../third_party/mmyolo/configs/yolov8/yolov8_m_mask-refine_syncbn_fast_8xb16-500e_coco.py'
)
custom_imports = dict(imports=['yolo_world'], allow_failed_imports=False)
# hyper-parameters
num_classes = 1203
num_training_classes = 80
max_epochs = 80 # Maximum training epochs
close_mosaic_epochs = 10
save_epoch_intervals = 5
text_channels = 512
neck_embed_channels = [128, 256, _base_.last_stage_out_channels // 2]
neck_num_heads = [4, 8, _base_.last_stage_out_channels // 2 // 32]
base_lr = 2e-4
weight_decay = 0.05
train_batch_size_per_gpu = 8
load_from = 'pretrained_models/yolo_world_m_clip_base_dual_vlpan_2e-3adamw_32xb16_100e_o365_goldg_train_pretrained-2b7bd1be.pth'
persistent_workers = False
# Polygon2Mask
downsample_ratio = 4
mask_overlap = False
use_mask2refine = True
max_aspect_ratio = 100
min_area_ratio = 0.01
# model settings
model = dict(
type='YOLOWorldDetector',
mm_neck=True,
num_train_classes=num_training_classes,
num_test_classes=num_classes,
data_preprocessor=dict(type='YOLOWDetDataPreprocessor'),
backbone=dict(
_delete_=True,
type='MultiModalYOLOBackbone',
image_model={{_base_.model.backbone}},
text_model=dict(
type='HuggingCLIPLanguageBackbone',
model_name='openai/clip-vit-base-patch32',
frozen_modules=[])),
neck=dict(type='YOLOWorldDualPAFPN',
guide_channels=text_channels,
embed_channels=neck_embed_channels,
num_heads=neck_num_heads,
block_cfg=dict(type='MaxSigmoidCSPLayerWithTwoConv'),
text_enhancder=dict(type='ImagePoolingAttentionModule',
embed_channels=256,
num_heads=8)),
bbox_head=dict(type='YOLOWorldSegHead',
head_module=dict(type='YOLOWorldSegHeadModule',
embed_dims=text_channels,
num_classes=num_training_classes,
mask_channels=32,
proto_channels=256),
mask_overlap=mask_overlap,
loss_mask=dict(type='mmdet.CrossEntropyLoss',
use_sigmoid=True,
reduction='none'),
loss_mask_weight=1.0),
train_cfg=dict(assigner=dict(num_classes=num_training_classes)),
test_cfg=dict(mask_thr_binary=0.5, fast_test=True))
pre_transform = [
dict(type='LoadImageFromFile', backend_args=_base_.backend_args),
dict(type='LoadAnnotations',
with_bbox=True,
with_mask=True,
mask2bbox=True)
]
last_transform = [
dict(type='mmdet.Albu',
transforms=_base_.albu_train_transforms,
bbox_params=dict(type='BboxParams',
format='pascal_voc',
label_fields=['gt_bboxes_labels',
'gt_ignore_flags']),
keymap={
'img': 'image',
'gt_bboxes': 'bboxes'
}),
dict(type='YOLOv5HSVRandomAug'),
dict(type='mmdet.RandomFlip', prob=0.5),
dict(type='Polygon2Mask',
downsample_ratio=downsample_ratio,
mask_overlap=mask_overlap),
]
# dataset settings
text_transform = [
dict(type='RandomLoadText',
num_neg_samples=(num_classes, num_classes),
max_num_samples=num_training_classes,
padding_to_max=True,
padding_value=''),
dict(type='PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape', 'flip',
'flip_direction', 'texts'))
]
mosaic_affine_transform = [
dict(type='MultiModalMosaic',
img_scale=_base_.img_scale,
pad_val=114.0,
pre_transform=pre_transform),
dict(type='YOLOv5CopyPaste', prob=_base_.copypaste_prob),
dict(
type='YOLOv5RandomAffine',
max_rotate_degree=0.0,
max_shear_degree=0.0,
max_aspect_ratio=100.,
scaling_ratio_range=(1 - _base_.affine_scale, 1 + _base_.affine_scale),
# img_scale is (width, height)
border=(-_base_.img_scale[0] // 2, -_base_.img_scale[1] // 2),
border_val=(114, 114, 114),
min_area_ratio=_base_.min_area_ratio,
use_mask_refine=True)
]
train_pipeline = [
*pre_transform, *mosaic_affine_transform,
dict(type='YOLOv5MultiModalMixUp',
prob=_base_.mixup_prob,
pre_transform=[*pre_transform, *mosaic_affine_transform]),
*last_transform, *text_transform
]
_train_pipeline_stage2 = [
*pre_transform,
dict(type='YOLOv5KeepRatioResize', scale=_base_.img_scale),
dict(type='LetterResize',
scale=_base_.img_scale,
allow_scale_up=True,
pad_val=dict(img=114.0)),
dict(type='YOLOv5RandomAffine',
max_rotate_degree=0.0,
max_shear_degree=0.0,
scaling_ratio_range=(1 - _base_.affine_scale,
1 + _base_.affine_scale),
max_aspect_ratio=_base_.max_aspect_ratio,
border_val=(114, 114, 114),
min_area_ratio=min_area_ratio,
use_mask_refine=use_mask2refine), *last_transform
]
train_pipeline_stage2 = [*_train_pipeline_stage2, *text_transform]
coco_train_dataset = dict(
_delete_=True,
type='MultiModalDataset',
dataset=dict(type='YOLOv5LVISV1Dataset',
data_root='data/coco',
ann_file='lvis/lvis_v1_train_base.json',
data_prefix=dict(img=''),
filter_cfg=dict(filter_empty_gt=True, min_size=32)),
class_text_path='data/texts/lvis_v1_base_class_texts.json',
pipeline=train_pipeline)
train_dataloader = dict(persistent_workers=persistent_workers,
batch_size=train_batch_size_per_gpu,
collate_fn=dict(type='yolow_collate'),
dataset=coco_train_dataset)
test_pipeline = [
*_base_.test_pipeline[:-1],
dict(type='LoadText'),
dict(type='mmdet.PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape',
'scale_factor', 'pad_param', 'texts'))
]
# training settings
default_hooks = dict(param_scheduler=dict(scheduler_type='linear',
lr_factor=0.01,
max_epochs=max_epochs),
checkpoint=dict(max_keep_ckpts=-1,
save_best=None,
interval=save_epoch_intervals))
custom_hooks = [
dict(type='EMAHook',
ema_type='ExpMomentumEMA',
momentum=0.0001,
update_buffers=True,
strict_load=False,
priority=49),
dict(type='mmdet.PipelineSwitchHook',
switch_epoch=max_epochs - close_mosaic_epochs,
switch_pipeline=train_pipeline_stage2)
]
train_cfg = dict(max_epochs=max_epochs,
val_interval=5,
dynamic_intervals=[((max_epochs - close_mosaic_epochs),
_base_.val_interval_stage2)])
optim_wrapper = dict(optimizer=dict(
_delete_=True,
type='AdamW',
lr=base_lr,
weight_decay=weight_decay,
batch_size_per_gpu=train_batch_size_per_gpu),
paramwise_cfg=dict(bias_decay_mult=0.0,
norm_decay_mult=0.0,
custom_keys={
'backbone.text_model':
dict(lr_mult=0.01),
'logit_scale':
dict(weight_decay=0.0)
}),
constructor='YOLOWv5OptimizerConstructor')
# evaluation settings
coco_val_dataset = dict(
_delete_=True,
type='MultiModalDataset',
dataset=dict(type='YOLOv5LVISV1Dataset',
data_root='data/coco/',
test_mode=True,
ann_file='lvis/lvis_v1_val.json',
data_prefix=dict(img=''),
batch_shapes_cfg=None),
class_text_path='data/captions/lvis_v1_class_captions.json',
pipeline=test_pipeline)
val_dataloader = dict(dataset=coco_val_dataset)
test_dataloader = val_dataloader
val_evaluator = dict(type='mmdet.LVISMetric',
ann_file='data/coco/lvis/lvis_v1_val.json',
metric=['bbox', 'segm'])
test_evaluator = val_evaluator
find_unused_parameters = True
configs/segmentation/yolo_world_seg_m_dual_vlpan_2e-4_80e_8gpus_seghead_finetune_lvis.py 0 → 100644
View file @ e9cee049
_base_ = (
'../../third_party/mmyolo/configs/yolov8/yolov8_m_mask-refine_syncbn_fast_8xb16-500e_coco.py'
)
custom_imports = dict(imports=['yolo_world'], allow_failed_imports=False)
# hyper-parameters
num_classes = 1203
num_training_classes = 80
max_epochs = 80 # Maximum training epochs
close_mosaic_epochs = 10
save_epoch_intervals = 5
text_channels = 512
neck_embed_channels = [128, 256, _base_.last_stage_out_channels // 2]
neck_num_heads = [4, 8, _base_.last_stage_out_channels // 2 // 32]
base_lr = 2e-4
weight_decay = 0.05
train_batch_size_per_gpu = 8
load_from = 'pretrained_models/yolo_world_m_clip_base_dual_vlpan_2e-3adamw_32xb16_100e_o365_goldg_train_pretrained-2b7bd1be.pth'
persistent_workers = False
# Polygon2Mask
downsample_ratio = 4
mask_overlap = False
use_mask2refine = True
max_aspect_ratio = 100
min_area_ratio = 0.01
# model settings
model = dict(
type='YOLOWorldDetector',
mm_neck=True,
num_train_classes=num_training_classes,
num_test_classes=num_classes,
data_preprocessor=dict(type='YOLOWDetDataPreprocessor'),
backbone=dict(
_delete_=True,
type='MultiModalYOLOBackbone',
image_model={{_base_.model.backbone}},
frozen_stages=4, # frozen the image backbone
text_model=dict(
type='HuggingCLIPLanguageBackbone',
model_name='openai/clip-vit-base-patch32',
frozen_modules=['all'])),
neck=dict(type='YOLOWorldDualPAFPN',
freeze_all=True,
guide_channels=text_channels,
embed_channels=neck_embed_channels,
num_heads=neck_num_heads,
block_cfg=dict(type='MaxSigmoidCSPLayerWithTwoConv'),
text_enhancder=dict(type='ImagePoolingAttentionModule',
embed_channels=256,
num_heads=8)),
bbox_head=dict(type='YOLOWorldSegHead',
head_module=dict(type='YOLOWorldSegHeadModule',
embed_dims=text_channels,
num_classes=num_training_classes,
mask_channels=32,
proto_channels=256,
freeze_bbox=True),
mask_overlap=mask_overlap,
loss_mask=dict(type='mmdet.CrossEntropyLoss',
use_sigmoid=True,
reduction='none'),
loss_mask_weight=1.0),
train_cfg=dict(assigner=dict(num_classes=num_training_classes)),
test_cfg=dict(mask_thr_binary=0.5, fast_test=True))
pre_transform = [
dict(type='LoadImageFromFile', backend_args=_base_.backend_args),
dict(type='LoadAnnotations',
with_bbox=True,
with_mask=True,
mask2bbox=True)
]
last_transform = [
dict(type='mmdet.Albu',
transforms=_base_.albu_train_transforms,
bbox_params=dict(type='BboxParams',
format='pascal_voc',
label_fields=['gt_bboxes_labels',
'gt_ignore_flags']),
keymap={
'img': 'image',
'gt_bboxes': 'bboxes'
}),
dict(type='YOLOv5HSVRandomAug'),
dict(type='mmdet.RandomFlip', prob=0.5),
dict(type='Polygon2Mask',
downsample_ratio=downsample_ratio,
mask_overlap=mask_overlap),
]
# dataset settings
text_transform = [
dict(type='RandomLoadText',
num_neg_samples=(num_classes, num_classes),
max_num_samples=num_training_classes,
padding_to_max=True,
padding_value=''),
dict(type='PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape', 'flip',
'flip_direction', 'texts'))
]
mosaic_affine_transform = [
dict(type='MultiModalMosaic',
img_scale=_base_.img_scale,
pad_val=114.0,
pre_transform=pre_transform),
dict(type='YOLOv5CopyPaste', prob=_base_.copypaste_prob),
dict(
type='YOLOv5RandomAffine',
max_rotate_degree=0.0,
max_shear_degree=0.0,
max_aspect_ratio=100.,
scaling_ratio_range=(1 - _base_.affine_scale, 1 + _base_.affine_scale),
# img_scale is (width, height)
border=(-_base_.img_scale[0] // 2, -_base_.img_scale[1] // 2),
border_val=(114, 114, 114),
min_area_ratio=_base_.min_area_ratio,
use_mask_refine=True)
]
train_pipeline = [
*pre_transform, *mosaic_affine_transform,
dict(type='YOLOv5MultiModalMixUp',
prob=_base_.mixup_prob,
pre_transform=[*pre_transform, *mosaic_affine_transform]),
*last_transform, *text_transform
]
_train_pipeline_stage2 = [
*pre_transform,
dict(type='YOLOv5KeepRatioResize', scale=_base_.img_scale),
dict(type='LetterResize',
scale=_base_.img_scale,
allow_scale_up=True,
pad_val=dict(img=114.0)),
dict(type='YOLOv5RandomAffine',
max_rotate_degree=0.0,
max_shear_degree=0.0,
scaling_ratio_range=(1 - _base_.affine_scale,
1 + _base_.affine_scale),
max_aspect_ratio=_base_.max_aspect_ratio,
border_val=(114, 114, 114),
min_area_ratio=min_area_ratio,
use_mask_refine=use_mask2refine), *last_transform
]
train_pipeline_stage2 = [*_train_pipeline_stage2, *text_transform]
coco_train_dataset = dict(
_delete_=True,
type='MultiModalDataset',
dataset=dict(type='YOLOv5LVISV1Dataset',
data_root='data/coco',
ann_file='lvis/lvis_v1_train_base.json',
data_prefix=dict(img=''),
filter_cfg=dict(filter_empty_gt=True, min_size=32)),
class_text_path='data/texts/lvis_v1_base_class_texts.json',
pipeline=train_pipeline)
train_dataloader = dict(persistent_workers=persistent_workers,
batch_size=train_batch_size_per_gpu,
collate_fn=dict(type='yolow_collate'),
dataset=coco_train_dataset)
test_pipeline = [
*_base_.test_pipeline[:-1],
dict(type='LoadText'),
dict(type='mmdet.PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape',
'scale_factor', 'pad_param', 'texts'))
]
# training settings
default_hooks = dict(param_scheduler=dict(scheduler_type='linear',
lr_factor=0.01,
max_epochs=max_epochs),
checkpoint=dict(max_keep_ckpts=-1,
save_best=None,
interval=save_epoch_intervals))
custom_hooks = [
dict(type='EMAHook',
ema_type='ExpMomentumEMA',
momentum=0.0001,
update_buffers=True,
strict_load=False,
priority=49),
dict(type='mmdet.PipelineSwitchHook',
switch_epoch=max_epochs - close_mosaic_epochs,
switch_pipeline=train_pipeline_stage2)
]
train_cfg = dict(max_epochs=max_epochs,
val_interval=5,
dynamic_intervals=[((max_epochs - close_mosaic_epochs),
_base_.val_interval_stage2)])
optim_wrapper = dict(optimizer=dict(
_delete_=True,
type='AdamW',
lr=base_lr,
weight_decay=weight_decay,
batch_size_per_gpu=train_batch_size_per_gpu),
paramwise_cfg=dict(bias_decay_mult=0.0,
norm_decay_mult=0.0,
custom_keys={
'backbone.text_model':
dict(lr_mult=0.01),
'logit_scale':
dict(weight_decay=0.0),
'neck':
dict(lr_mult=0.0),
'head.head_module.reg_preds':
dict(lr_mult=0.0),
'head.head_module.cls_preds':
dict(lr_mult=0.0),
'head.head_module.cls_contrasts':
dict(lr_mult=0.0)
}),
constructor='YOLOWv5OptimizerConstructor')
# evaluation settings
coco_val_dataset = dict(
_delete_=True,
type='MultiModalDataset',
dataset=dict(type='YOLOv5LVISV1Dataset',
data_root='data/coco/',
test_mode=True,
ann_file='lvis/lvis_v1_val.json',
data_prefix=dict(img=''),
batch_shapes_cfg=None),
class_text_path='data/captions/lvis_v1_class_captions.json',
pipeline=test_pipeline)
val_dataloader = dict(dataset=coco_val_dataset)
test_dataloader = val_dataloader
val_evaluator = dict(type='mmdet.LVISMetric',
ann_file='data/coco/lvis/lvis_v1_val.json',
metric=['bbox', 'segm'])
test_evaluator = val_evaluator
find_unused_parameters = True
configs/segmentation/yolo_world_v2_seg_l_vlpan_bn_2e-4_80e_8gpus_seghead_finetune_lvis.py 0 → 100644
View file @ e9cee049
_base_ = (
'../../third_party/mmyolo/configs/yolov8/yolov8_l_mask-refine_syncbn_fast_8xb16-500e_coco.py'
)
custom_imports = dict(imports=['yolo_world'], allow_failed_imports=False)
# hyper-parameters
num_classes = 1203
num_training_classes = 80
max_epochs = 80 # Maximum training epochs
close_mosaic_epochs = 10
save_epoch_intervals = 5
text_channels = 512
neck_embed_channels = [128, 256, _base_.last_stage_out_channels // 2]
neck_num_heads = [4, 8, _base_.last_stage_out_channels // 2 // 32]
base_lr = 2e-4
weight_decay = 0.05
train_batch_size_per_gpu = 8
load_from = 'pretrained_models/yolo_world_l_clip_t2i_bn_2e-3adamw_32xb16-100e_obj365v1_goldg_cc3mlite_train-ca93cd1f.pth'
# text_model_name = '../pretrained_models/clip-vit-base-patch32-projection'
text_model_name = 'openai/clip-vit-base-patch32'
persistent_workers = False
# Polygon2Mask
downsample_ratio = 4
mask_overlap = False
use_mask2refine = True
max_aspect_ratio = 100
min_area_ratio = 0.01
# model settings
model = dict(
type='YOLOWorldDetector',
mm_neck=True,
num_train_classes=num_training_classes,
num_test_classes=num_classes,
data_preprocessor=dict(type='YOLOWDetDataPreprocessor'),
backbone=dict(
_delete_=True,
type='MultiModalYOLOBackbone',
image_model={{_base_.model.backbone}},
frozen_stages=4, # frozen the image backbone
text_model=dict(
type='HuggingCLIPLanguageBackbone',
model_name=text_model_name,
frozen_modules=['all'])),
neck=dict(type='YOLOWorldPAFPN',
freeze_all=True,
guide_channels=text_channels,
embed_channels=neck_embed_channels,
num_heads=neck_num_heads,
block_cfg=dict(type='MaxSigmoidCSPLayerWithTwoConv')),
bbox_head=dict(type='YOLOWorldSegHead',
head_module=dict(type='YOLOWorldSegHeadModule',
use_bn_head=True,
embed_dims=text_channels,
num_classes=num_training_classes,
mask_channels=32,
proto_channels=256,
freeze_bbox=True),
mask_overlap=mask_overlap,
loss_mask=dict(type='mmdet.CrossEntropyLoss',
use_sigmoid=True,
reduction='none'),
loss_mask_weight=1.0),
train_cfg=dict(assigner=dict(
type='YOLOWorldSegAssigner',
num_classes=num_training_classes)),
test_cfg=dict(mask_thr_binary=0.5, fast_test=True))
pre_transform = [
dict(type='LoadImageFromFile', backend_args=_base_.backend_args),
dict(type='LoadAnnotations',
with_bbox=True,
with_mask=True,
mask2bbox=True)
]
last_transform = [
dict(type='mmdet.Albu',
transforms=_base_.albu_train_transforms,
bbox_params=dict(type='BboxParams',
format='pascal_voc',
label_fields=['gt_bboxes_labels',
'gt_ignore_flags']),
keymap={
'img': 'image',
'gt_bboxes': 'bboxes'
}),
dict(type='YOLOv5HSVRandomAug'),
dict(type='mmdet.RandomFlip', prob=0.5),
dict(type='Polygon2Mask',
downsample_ratio=downsample_ratio,
mask_overlap=mask_overlap),
]
# dataset settings
text_transform = [
dict(type='RandomLoadText',
num_neg_samples=(num_classes, num_classes),
max_num_samples=num_training_classes,
padding_to_max=True,
padding_value=''),
dict(type='PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape', 'flip',
'flip_direction', 'texts'))
]
mosaic_affine_transform = [
dict(type='MultiModalMosaic',
img_scale=_base_.img_scale,
pad_val=114.0,
pre_transform=pre_transform),
dict(type='YOLOv5CopyPaste', prob=_base_.copypaste_prob),
dict(
type='YOLOv5RandomAffine',
max_rotate_degree=0.0,
max_shear_degree=0.0,
max_aspect_ratio=100.,
scaling_ratio_range=(1 - _base_.affine_scale, 1 + _base_.affine_scale),
# img_scale is (width, height)
border=(-_base_.img_scale[0] // 2, -_base_.img_scale[1] // 2),
border_val=(114, 114, 114),
min_area_ratio=_base_.min_area_ratio,
use_mask_refine=True)
]
train_pipeline = [
*pre_transform, *mosaic_affine_transform,
dict(type='YOLOv5MultiModalMixUp',
prob=_base_.mixup_prob,
pre_transform=[*pre_transform, *mosaic_affine_transform]),
*last_transform, *text_transform
]
_train_pipeline_stage2 = [
*pre_transform,
dict(type='YOLOv5KeepRatioResize', scale=_base_.img_scale),
dict(type='LetterResize',
scale=_base_.img_scale,
allow_scale_up=True,
pad_val=dict(img=114.0)),
dict(type='YOLOv5RandomAffine',
max_rotate_degree=0.0,
max_shear_degree=0.0,
scaling_ratio_range=(1 - _base_.affine_scale,
1 + _base_.affine_scale),
max_aspect_ratio=_base_.max_aspect_ratio,
border_val=(114, 114, 114),
min_area_ratio=min_area_ratio,
use_mask_refine=use_mask2refine), *last_transform
]
train_pipeline_stage2 = [*_train_pipeline_stage2, *text_transform]
coco_train_dataset = dict(
_delete_=True,
type='MultiModalDataset',
dataset=dict(type='YOLOv5LVISV1Dataset',
data_root='data/coco',
ann_file='lvis/lvis_v1_train_base.json',
data_prefix=dict(img=''),
filter_cfg=dict(filter_empty_gt=True, min_size=32)),
class_text_path='data/texts/lvis_v1_base_class_texts.json',
pipeline=train_pipeline)
train_dataloader = dict(persistent_workers=persistent_workers,
batch_size=train_batch_size_per_gpu,
collate_fn=dict(type='yolow_collate'),
dataset=coco_train_dataset)
test_pipeline = [
*_base_.test_pipeline[:-1],
dict(type='LoadText'),
dict(type='mmdet.PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape',
'scale_factor', 'pad_param', 'texts'))
]
# training settings
default_hooks = dict(param_scheduler=dict(scheduler_type='linear',
lr_factor=0.01,
max_epochs=max_epochs),
checkpoint=dict(max_keep_ckpts=-1,
save_best=None,
interval=save_epoch_intervals))
custom_hooks = [
dict(type='EMAHook',
ema_type='ExpMomentumEMA',
momentum=0.0001,
update_buffers=True,
strict_load=False,
priority=49),
dict(type='mmdet.PipelineSwitchHook',
switch_epoch=max_epochs - close_mosaic_epochs,
switch_pipeline=train_pipeline_stage2)
]
train_cfg = dict(max_epochs=max_epochs,
val_interval=5,
dynamic_intervals=[((max_epochs - close_mosaic_epochs),
_base_.val_interval_stage2)])
optim_wrapper = dict(optimizer=dict(
_delete_=True,
type='AdamW',
lr=base_lr,
weight_decay=weight_decay,
batch_size_per_gpu=train_batch_size_per_gpu),
paramwise_cfg=dict(bias_decay_mult=0.0,
norm_decay_mult=0.0,
custom_keys={
'backbone.text_model':
dict(lr_mult=0.01),
'logit_scale':
dict(weight_decay=0.0),
'neck':
dict(lr_mult=0.0),
'head.head_module.reg_preds':
dict(lr_mult=0.0),
'head.head_module.cls_preds':
dict(lr_mult=0.0),
'head.head_module.cls_contrasts':
dict(lr_mult=0.0)
}),
constructor='YOLOWv5OptimizerConstructor')
# evaluation settings
coco_val_dataset = dict(
_delete_=True,
type='MultiModalDataset',
dataset=dict(type='YOLOv5LVISV1Dataset',
data_root='data/coco/',
test_mode=True,
ann_file='lvis/lvis_v1_val.json',
data_prefix=dict(img=''),
batch_shapes_cfg=None),
class_text_path='data/texts/lvis_v1_class_texts.json',
pipeline=test_pipeline)
val_dataloader = dict(dataset=coco_val_dataset)
test_dataloader = val_dataloader
val_evaluator = dict(type='mmdet.LVISMetric',
ann_file='data/coco/lvis/lvis_v1_val.json',
metric=['bbox', 'segm'])
test_evaluator = val_evaluator
find_unused_parameters = True
configs/segmentation/yolo_world_v2_seg_m_vlpan_bn_2e-4_80e_8gpus_seghead_finetune_lvis.py 0 → 100644
View file @ e9cee049
_base_ = (
'../../third_party/mmyolo/configs/yolov8/yolov8_m_mask-refine_syncbn_fast_8xb16-500e_coco.py'
)
custom_imports = dict(imports=['yolo_world'], allow_failed_imports=False)
# hyper-parameters
num_classes = 1203
num_training_classes = 80
max_epochs = 80 # Maximum training epochs
close_mosaic_epochs = 10
save_epoch_intervals = 5
text_channels = 512
neck_embed_channels = [128, 256, _base_.last_stage_out_channels // 2]
neck_num_heads = [4, 8, _base_.last_stage_out_channels // 2 // 32]
base_lr = 2e-4
weight_decay = 0.05
train_batch_size_per_gpu = 8
load_from = 'pretrained_models/yolo_world_m_clip_t2i_bn_2e-3adamw_32xb16-100e_obj365v1_goldg_train-c6237d5b.pth'
text_model_name = '../pretrained_models/clip-vit-base-patch32-projection'
# text_model_name = 'openai/clip-vit-base-patch32'
persistent_workers = False
# Polygon2Mask
downsample_ratio = 4
mask_overlap = False
use_mask2refine = True
max_aspect_ratio = 100
min_area_ratio = 0.01
# model settings
model = dict(
type='YOLOWorldDetector',
mm_neck=True,
num_train_classes=num_training_classes,
num_test_classes=num_classes,
data_preprocessor=dict(type='YOLOWDetDataPreprocessor'),
backbone=dict(
_delete_=True,
type='MultiModalYOLOBackbone',
image_model={{_base_.model.backbone}},
frozen_stages=4, # frozen the image backbone
text_model=dict(
type='HuggingCLIPLanguageBackbone',
model_name=text_model_name,
frozen_modules=['all'])),
neck=dict(type='YOLOWorldPAFPN',
freeze_all=True,
guide_channels=text_channels,
embed_channels=neck_embed_channels,
num_heads=neck_num_heads,
block_cfg=dict(type='MaxSigmoidCSPLayerWithTwoConv')),
bbox_head=dict(type='YOLOWorldSegHead',
head_module=dict(type='YOLOWorldSegHeadModule',
use_bn_head=True,
embed_dims=text_channels,
num_classes=num_training_classes,
mask_channels=32,
proto_channels=256,
freeze_bbox=True),
mask_overlap=mask_overlap,
loss_mask=dict(type='mmdet.CrossEntropyLoss',
use_sigmoid=True,
reduction='none'),
loss_mask_weight=1.0),
train_cfg=dict(assigner=dict(
type='YOLOWorldSegAssigner',
num_classes=num_training_classes)),
test_cfg=dict(mask_thr_binary=0.5, fast_test=True))
pre_transform = [
dict(type='LoadImageFromFile', backend_args=_base_.backend_args),
dict(type='LoadAnnotations',
with_bbox=True,
with_mask=True,
mask2bbox=True)
]
last_transform = [
dict(type='mmdet.Albu',
transforms=_base_.albu_train_transforms,
bbox_params=dict(type='BboxParams',
format='pascal_voc',
label_fields=['gt_bboxes_labels',
'gt_ignore_flags']),
keymap={
'img': 'image',
'gt_bboxes': 'bboxes'
}),
dict(type='YOLOv5HSVRandomAug'),
dict(type='mmdet.RandomFlip', prob=0.5),
dict(type='Polygon2Mask',
downsample_ratio=downsample_ratio,
mask_overlap=mask_overlap),
]
# dataset settings
text_transform = [
dict(type='RandomLoadText',
num_neg_samples=(num_classes, num_classes),
max_num_samples=num_training_classes,
padding_to_max=True,
padding_value=''),
dict(type='PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape', 'flip',
'flip_direction', 'texts'))
]
mosaic_affine_transform = [
dict(type='MultiModalMosaic',
img_scale=_base_.img_scale,
pad_val=114.0,
pre_transform=pre_transform),
dict(type='YOLOv5CopyPaste', prob=_base_.copypaste_prob),
dict(
type='YOLOv5RandomAffine',
max_rotate_degree=0.0,
max_shear_degree=0.0,
max_aspect_ratio=100.,
scaling_ratio_range=(1 - _base_.affine_scale, 1 + _base_.affine_scale),
# img_scale is (width, height)
border=(-_base_.img_scale[0] // 2, -_base_.img_scale[1] // 2),
border_val=(114, 114, 114),
min_area_ratio=_base_.min_area_ratio,
use_mask_refine=True)
]
train_pipeline = [
*pre_transform, *mosaic_affine_transform,
dict(type='YOLOv5MultiModalMixUp',
prob=_base_.mixup_prob,
pre_transform=[*pre_transform, *mosaic_affine_transform]),
*last_transform, *text_transform
]
_train_pipeline_stage2 = [
*pre_transform,
dict(type='YOLOv5KeepRatioResize', scale=_base_.img_scale),
dict(type='LetterResize',
scale=_base_.img_scale,
allow_scale_up=True,
pad_val=dict(img=114.0)),
dict(type='YOLOv5RandomAffine',
max_rotate_degree=0.0,
max_shear_degree=0.0,
scaling_ratio_range=(1 - _base_.affine_scale,
1 + _base_.affine_scale),
max_aspect_ratio=_base_.max_aspect_ratio,
border_val=(114, 114, 114),
min_area_ratio=min_area_ratio,
use_mask_refine=use_mask2refine), *last_transform
]
train_pipeline_stage2 = [*_train_pipeline_stage2, *text_transform]
coco_train_dataset = dict(
_delete_=True,
type='MultiModalDataset',
dataset=dict(type='YOLOv5LVISV1Dataset',
data_root='data/coco',
ann_file='lvis/lvis_v1_train_base.json',
data_prefix=dict(img=''),
filter_cfg=dict(filter_empty_gt=True, min_size=32)),
class_text_path='data/texts/lvis_v1_base_class_texts.json',
pipeline=train_pipeline)
train_dataloader = dict(persistent_workers=persistent_workers,
batch_size=train_batch_size_per_gpu,
collate_fn=dict(type='yolow_collate'),
dataset=coco_train_dataset)
test_pipeline = [
*_base_.test_pipeline[:-1],
dict(type='LoadText'),
dict(type='mmdet.PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape',
'scale_factor', 'pad_param', 'texts'))
]
# training settings
default_hooks = dict(param_scheduler=dict(scheduler_type='linear',
lr_factor=0.01,
max_epochs=max_epochs),
checkpoint=dict(max_keep_ckpts=-1,
save_best=None,
interval=save_epoch_intervals))
custom_hooks = [
dict(type='EMAHook',
ema_type='ExpMomentumEMA',
momentum=0.0001,
update_buffers=True,
strict_load=False,
priority=49),
dict(type='mmdet.PipelineSwitchHook',
switch_epoch=max_epochs - close_mosaic_epochs,
switch_pipeline=train_pipeline_stage2)
]
train_cfg = dict(max_epochs=max_epochs,
val_interval=5,
dynamic_intervals=[((max_epochs - close_mosaic_epochs),
_base_.val_interval_stage2)])
optim_wrapper = dict(optimizer=dict(
_delete_=True,
type='AdamW',
lr=base_lr,
weight_decay=weight_decay,
batch_size_per_gpu=train_batch_size_per_gpu),
paramwise_cfg=dict(bias_decay_mult=0.0,
norm_decay_mult=0.0,
custom_keys={
'backbone.text_model':
dict(lr_mult=0.01),
'logit_scale':
dict(weight_decay=0.0),
'neck':
dict(lr_mult=0.0),
'head.head_module.reg_preds':
dict(lr_mult=0.0),
'head.head_module.cls_preds':
dict(lr_mult=0.0),
'head.head_module.cls_contrasts':
dict(lr_mult=0.0)
}),
constructor='YOLOWv5OptimizerConstructor')
# evaluation settings
coco_val_dataset = dict(
_delete_=True,
type='MultiModalDataset',
dataset=dict(type='YOLOv5LVISV1Dataset',
data_root='data/coco/',
test_mode=True,
ann_file='lvis/lvis_v1_val.json',
data_prefix=dict(img=''),
batch_shapes_cfg=None),
class_text_path='data/texts/lvis_v1_class_texts.json',
pipeline=test_pipeline)
val_dataloader = dict(dataset=coco_val_dataset)
test_dataloader = val_dataloader
val_evaluator = dict(type='mmdet.LVISMetric',
ann_file='data/coco/lvis/lvis_v1_val.json',
metric=['bbox', 'segm'])
test_evaluator = val_evaluator
find_unused_parameters = True
data/texts/coco_class_texts.json 0 → 100644
View file @ e9cee049
[["person"], ["bicycle"], ["car"], ["motorcycle"], ["airplane"], ["bus"], ["train"], ["truck"], ["boat"], ["traffic light"], ["fire hydrant"], ["stop sign"], ["parking meter"], ["bench"], ["bird"], ["cat"], ["dog"], ["horse"], ["sheep"], ["cow"], ["elephant"], ["bear"], ["zebra"], ["giraffe"], ["backpack"], ["umbrella"], ["handbag"], ["tie"], ["suitcase"], ["frisbee"], ["skis"], ["snowboard"], ["sports ball"], ["kite"], ["baseball bat"], ["baseball glove"], ["skateboard"], ["surfboard"], ["tennis racket"], ["bottle"], ["wine glass"], ["cup"], ["fork"], ["knife"], ["spoon"], ["bowl"], ["banana"], ["apple"], ["sandwich"], ["orange"], ["broccoli"], ["carrot"], ["hot dog"], ["pizza"], ["donut"], ["cake"], ["chair"], ["couch"], ["potted plant"], ["bed"], ["dining table"], ["toilet"], ["tv"], ["laptop"], ["mouse"], ["remote"], ["keyboard"], ["cell phone"], ["microwave"], ["oven"], ["toaster"], ["sink"], ["refrigerator"], ["book"], ["clock"], ["vase"], ["scissors"], ["teddy bear"], ["hair drier"], ["toothbrush"]]
data/texts/lvis_v1_base_class_captions.json 0 → 100644
View file @ e9cee049
[["aerosol can", "spray can"], ["air conditioner"], ["airplane", "aeroplane"], ["alarm clock"], ["alcohol", "alcoholic beverage"], ["alligator", "gator"], ["almond"], ["ambulance"], ["amplifier"], ["anklet", "ankle bracelet"], ["antenna", "aerial", "transmitting aerial"], ["apple"], ["apron"], ["aquarium", "fish tank"], ["armband"], ["armchair"], ["artichoke"], ["trash can", "garbage can", "wastebin", "dustbin", "trash barrel", "trash bin"], ["ashtray"], ["asparagus"], ["atomizer", "atomiser", "spray", "sprayer", "nebulizer", "nebuliser"], ["avocado"], ["award", "accolade"], ["awning"], ["baby buggy", "baby carriage", "perambulator", "pram", "stroller"], ["basketball backboard"], ["backpack", "knapsack", "packsack", "rucksack", "haversack"], ["handbag", "purse", "pocketbook"], ["suitcase", "baggage", "luggage"], ["bagel", "beigel"], ["ball"], ["balloon"], ["bamboo"], ["banana"], ["Band Aid"], ["bandage"], ["bandanna", "bandana"], ["banner", "streamer"], ["barrel", "cask"], ["barrette"], ["barrow", "garden cart", "lawn cart", "wheelbarrow"], ["baseball base"], ["baseball"], ["baseball bat"], ["baseball cap", "jockey cap", "golf cap"], ["baseball glove", "baseball mitt"], ["basket", "handbasket"], ["basketball"], ["bat", "bat animal"], ["bath mat"], ["bath towel"], ["bathrobe"], ["bathtub", "bathing tub"], ["battery"], ["bead"], ["bean curd", "tofu"], ["beanbag"], ["beanie", "beany"], ["bear"], ["bed"], ["bedspread", "bedcover", "bed covering", "counterpane", "spread"], ["cow"], ["beef", "beef food", "boeuf", "boeuf food"], ["beer bottle"], ["beer can"], ["bell"], ["bell pepper", "capsicum"], ["belt"], ["belt buckle"], ["bench"], ["beret"], ["bib"], ["bicycle", "bike", "bike bicycle"], ["visor", "vizor"], ["billboard"], ["binder", "ring-binder"], ["binoculars", "field glasses", "opera glasses"], ["bird"], ["birdfeeder"], ["birdbath"], ["birdcage"], ["birdhouse"], ["birthday cake"], ["black sheep"], ["blackberry"], ["blackboard", "chalkboard"], ["blanket"], ["blazer", "sport jacket", "sport coat", "sports jacket", "sports coat"], ["blender", "liquidizer", "liquidiser"], ["blinker", "flasher"], ["blouse"], ["blueberry"], ["boat", "ship", "ship boat"], ["bobbin", "spool", "reel"], ["bobby pin", "hairgrip"], ["boiled egg", "coddled egg"], ["deadbolt"], ["bolt"], ["book"], ["bookcase"], ["booklet", "brochure", "leaflet", "pamphlet"], ["boot"], ["bottle"], ["bottle opener"], ["bouquet"], ["bow", "bow decorative ribbons"], ["bow-tie", "bowtie"], ["bowl"], ["bowler hat", "bowler", "derby hat", "derby", "plug hat"], ["box"], ["suspenders"], ["bracelet", "bangle"], ["brassiere", "bra", "bandeau"], ["bread-bin", "breadbox"], ["bread"], ["bridal gown", "wedding gown", "wedding dress"], ["briefcase"], ["broccoli"], ["broom"], ["brownie"], ["brussels sprouts"], ["bucket", "pail"], ["horned cow"], ["bulldog"], ["bullet train"], ["bulletin board", "notice board"], ["bullhorn", "megaphone"], ["bun", "roll"], ["bunk bed"], ["buoy"], ["bus", "bus vehicle", "autobus", "charabanc", "double-decker", "motorbus", "motorcoach"], ["business card"], ["butter"], ["butterfly"], ["button"], ["cab", "cab taxi", "taxi", "taxicab"], ["cabin car", "caboose"], ["cabinet"], ["cake"], ["calculator"], ["calendar"], ["calf"], ["camcorder"], ["camel"], ["camera"], ["camera lens"], ["camper", "camper vehicle", "camping bus", "motor home"], ["can", "tin can"], ["can opener", "tin opener"], ["candle", "candlestick"], ["candle holder"], ["candy cane"], ["walking cane"], ["canister", "cannister"], ["canoe"], ["cantaloup", "cantaloupe"], ["cap", "cap headwear"], ["bottle cap", "cap", "cap container lid"], ["cape"], ["cappuccino", "coffee cappuccino"], ["car", "car automobile", "auto", "auto automobile", "automobile"], ["railcar", "railcar part of a train", "railway car", "railway car part of a train", "railroad car", "railroad car part of a train"], ["identity card"], ["card"], ["cardigan"], ["horse carriage"], ["carrot"], ["tote bag"], ["cart"], ["carton"], ["cash register", "register", "register for cash transactions"], ["cast", "plaster cast", "plaster bandage"], ["cat"], ["cauliflower"], ["cayenne", "cayenne spice", "cayenne pepper", "cayenne pepper spice", "red pepper", "red pepper spice"], ["CD player"], ["celery"], ["cellular telephone", "cellular phone", "cellphone", "mobile phone", "smart phone"], ["chair"], ["chandelier"], ["cherry"], ["chicken", "chicken animal"], ["chickpea", "garbanzo"], ["chili", "chili vegetable", "chili pepper", "chili pepper vegetable", "chilli", "chilli vegetable", "chilly", "chilly vegetable", "chile", "chile vegetable"], ["crisp", "crisp potato chip", "potato chip"], ["chocolate bar"], ["chocolate cake"], ["choker", "collar", "neckband"], ["chopping board", "cutting board", "chopping block"], ["chopstick"], ["Christmas tree"], ["slide"], ["cigarette"], ["cigarette case", "cigarette pack"], ["cistern", "water tank"], ["clasp"], ["cleansing agent", "cleanser", "cleaner"], ["clip"], ["clipboard"], ["clock", "timepiece", "timekeeper"], ["clock tower"], ["clothes hamper", "laundry basket", "clothes basket"], ["clothespin", "clothes peg"], ["coaster"], ["coat"], ["coat hanger", "clothes hanger", "dress hanger"], ["coatrack", "hatrack"], ["cock", "rooster"], ["coconut", "cocoanut"], ["coffee maker", "coffee machine"], ["coffee table", "cocktail table"], ["coffeepot"], ["coin"], ["colander", "cullender"], ["coleslaw", "slaw"], ["pacifier", "teething ring"], ["computer keyboard", "keyboard", "keyboard computer"], ["condiment"], ["cone", "traffic cone"], ["control", "controller"], ["cookie", "cooky", "biscuit", "biscuit cookie"], ["cooler", "cooler for food", "ice chest"], ["cork", "cork bottle plug", "bottle cork"], ["corkscrew", "bottle screw"], ["edible corn", "corn", "maize"], ["cornet", "horn", "trumpet"], ["cornice", "valance", "valance board", "pelmet"], ["corset", "girdle"], ["costume"], ["cowbell"], ["cowboy hat", "ten-gallon hat"], ["crab", "crab animal"], ["cracker"], ["crate"], ["crayon", "wax crayon"], ["crescent roll", "croissant"], ["crib", "cot"], ["crock pot", "earthenware jar"], ["crossbar"], ["crow"], ["crown"], ["crucifix"], ["cruise ship", "cruise liner"], ["police cruiser", "patrol car", "police car", "squad car"], ["crumb"], ["crutch"], ["cub", "cub animal"], ["cube", "square block"], ["cucumber", "cuke"], ["cufflink"], ["cup"], ["trophy cup"], ["cupboard", "closet"], ["cupcake"], ["curtain", "drapery"], ["cushion"], ["dartboard"], ["deck chair", "beach chair"], ["deer", "cervid"], ["dental floss", "floss"], ["desk"], ["diaper"], ["dining table"], ["dish"], ["dish antenna"], ["dishrag", "dishcloth"], ["dishtowel", "tea towel"], ["dishwasher", "dishwashing machine"], ["dispenser"], ["Dixie cup", "paper cup"], ["dog"], ["dog collar"], ["doll"], ["dolphin"], ["domestic ass", "donkey"], ["doorknob", "doorhandle"], ["doormat", "welcome mat"], ["doughnut", "donut"], ["drawer"], ["underdrawers", "boxers", "boxershorts"], ["dress", "frock"], ["dress hat", "high hat", "opera hat", "silk hat", "top hat"], ["dress suit"], ["dresser"], ["drill"], ["drum", "drum musical instrument"], ["duck"], ["duckling"], ["duct tape"], ["duffel bag", "duffle bag", "duffel", "duffle"], ["dumpster"], ["eagle"], ["earphone", "earpiece", "headphone"], ["earring"], ["easel"], ["egg", "eggs"], ["egg yolk", "yolk", "yolk egg"], ["eggbeater", "eggwhisk"], ["eggplant", "aubergine"], ["refrigerator"], ["elephant"], ["elk", "moose"], ["envelope"], ["eraser"], ["fan"], ["faucet", "spigot", "tap"], ["Ferris wheel"], ["ferry", "ferryboat"], ["fighter jet", "fighter aircraft", "attack aircraft"], ["figurine"], ["file cabinet", "filing cabinet"], ["fire alarm", "smoke alarm"], ["fire engine", "fire truck"], ["fire extinguisher", "extinguisher"], ["fire hose"], ["fireplace"], ["fireplug", "fire hydrant", "hydrant"], ["fish"], ["fish", "fish food"], ["fishing rod", "fishing pole"], ["flag"], ["flagpole", "flagstaff"], ["flamingo"], ["flannel"], ["flap"], ["flashlight", "torch"], ["flip-flop", "flip-flop sandal"], ["flipper", "flipper footwear", "fin", "fin footwear"], ["flower arrangement", "floral arrangement"], ["flute glass", "champagne flute"], ["foal"], ["folding chair"], ["food processor"], ["football", "football American"], ["footstool", "footrest"], ["fork"], ["forklift"], ["freight car"], ["French toast"], ["freshener", "air freshener"], ["frisbee"], ["frog", "toad", "toad frog"], ["fruit juice"], ["frying pan", "frypan", "skillet"], ["garbage truck"], ["garden hose"], ["gargle", "mouthwash"], ["garlic", "ail"], ["gazelle"], ["gelatin", "jelly"], ["giant panda", "panda", "panda bear"], ["gift wrap"], ["ginger", "gingerroot"], ["giraffe"], ["cincture", "sash", "waistband", "waistcloth"], ["glass", "glass drink container", "drinking glass"], ["globe"], ["glove"], ["goat"], ["goggles"], ["golf club", "golf-club"], ["golfcart"], ["goose"], ["grape"], ["grater"], ["gravestone", "headstone", "tombstone"], ["green bean"], ["green onion", "spring onion", "scallion"], ["grill", "grille", "grillwork", "radiator grille"], ["grizzly", "grizzly bear"], ["grocery bag"], ["guitar"], ["gull", "seagull"], ["gun"], ["hairbrush"], ["hairnet"], ["hairpin"], ["ham", "jambon", "gammon"], ["hamburger", "beefburger", "burger"], ["hammer"], ["hammock"], ["hamster"], ["hair dryer"], ["hand towel", "face towel"], ["handcart", "pushcart", "hand truck"], ["handkerchief"], ["handle", "grip", "handgrip"], ["hat"], ["veil"], ["headband"], ["headboard"], ["headlight", "headlamp"], ["headscarf"], ["headstall", "headstall for horses", "headpiece", "headpiece for horses"], ["heart"], ["heater", "warmer"], ["helicopter"], ["helmet"], ["highchair", "feeding chair"], ["hinge"], ["hog", "pig"], ["home plate", "home plate baseball", "home base", "home base baseball"], ["honey"], ["fume hood", "exhaust hood"], ["hook"], ["horse"], ["hose", "hosepipe"], ["hot sauce"], ["hummingbird"], ["polar bear"], ["icecream"], ["ice maker"], ["igniter", "ignitor", "lighter"], ["iPod"], ["iron", "iron for clothing", "smoothing iron", "smoothing iron for clothing"], ["ironing board"], ["jacket"], ["jam"], ["jar"], ["jean", "blue jean", "denim"], ["jeep", "landrover"], ["jersey", "T-shirt", "tee shirt"], ["jet plane", "jet-propelled plane"], ["jewelry", "jewellery"], ["jumpsuit"], ["kayak"], ["kettle", "boiler"], ["key"], ["kilt"], ["kimono"], ["kitchen sink"], ["kite"], ["kitten", "kitty"], ["kiwi fruit"], ["knee pad"], ["knife"], ["knob"], ["ladder"], ["ladle"], ["ladybug", "ladybeetle", "ladybird beetle"], ["lamb", "lamb animal"], ["lamp"], ["lamppost"], ["lampshade"], ["lantern"], ["lanyard", "laniard"], ["laptop computer", "notebook computer"], ["latch"], ["legging", "legging clothing", "leging", "leging clothing", "leg covering"], ["Lego", "Lego set"], ["lemon"], ["lettuce"], ["license plate", "numberplate"], ["life buoy", "lifesaver", "life belt", "life ring"], ["life jacket", "life vest"], ["lightbulb"], ["lime"], ["lion"], ["lip balm"], ["lizard"], ["log"], ["lollipop"], ["speaker", "speaker stereo equipment"], ["loveseat"], ["magazine"], ["magnet"], ["mail slot"], ["mailbox", "mailbox at home", "letter box", "letter box at home"], ["mandarin orange"], ["manger", "trough"], ["manhole"], ["map"], ["marker"], ["mashed potato"], ["mask", "facemask"], ["mast"], ["mat", "mat gym equipment", "gym mat"], ["mattress"], ["measuring cup"], ["measuring stick", "ruler", "ruler measuring stick", "measuring rod"], ["meatball"], ["medicine"], ["melon"], ["microphone"], ["microwave oven"], ["milk"], ["minivan"], ["mirror"], ["mitten"], ["mixer", "mixer kitchen tool", "stand mixer"], ["money"], ["monitor", "monitor computer equipment"], ["monkey"], ["motor"], ["motor scooter", "scooter"], ["motorcycle"], ["mound", "mound baseball", "pitcher's mound"], ["mouse", "mouse computer equipment", "computer mouse"], ["mousepad"], ["muffin"], ["mug"], ["mushroom"], ["musical instrument", "instrument", "instrument musical"], ["napkin", "table napkin", "serviette"], ["necklace"], ["necktie", "tie", "tie necktie"], ["needle"], ["nest"], ["newspaper", "paper", "paper newspaper"], ["newsstand"], ["nightshirt", "nightwear", "sleepwear", "nightclothes"], ["noseband", "noseband for animals", "nosepiece", "nosepiece for animals"], ["notebook"], ["notepad"], ["nut"], ["oar"], ["oil lamp", "kerosene lamp", "kerosine lamp"], ["olive oil"], ["onion"], ["orange", "orange fruit"], ["orange juice"], ["ostrich"], ["ottoman", "pouf", "pouffe", "hassock"], ["oven"], ["overalls", "overalls clothing"], ["owl"], ["packet"], ["pad"], ["paddle", "boat paddle"], ["padlock"], ["paintbrush"], ["painting"], ["pajamas", "pyjamas"], ["palette", "pallet"], ["pan", "pan for cooking", "cooking pan"], ["pancake"], ["paper plate"], ["paper towel"], ["parachute"], ["parakeet", "parrakeet", "parroket", "paraquet", "paroquet", "parroquet"], ["parasail", "parasail sports"], ["parasol", "sunshade"], ["parka", "anorak"], ["parking meter"], ["parrot"], ["passenger car", "passenger car part of a train", "coach", "coach part of a train"], ["passport"], ["pastry"], ["pea", "pea food"], ["peach"], ["peanut butter"], ["pear"], ["peeler", "peeler tool for fruit and vegetables"], ["pelican"], ["pen"], ["pencil"], ["penguin"], ["pepper", "peppercorn"], ["pepper mill", "pepper grinder"], ["perfume"], ["person", "baby", "child", "boy", "girl", "man", "woman", "human"], ["pet"], ["pew", "pew church bench", "church bench"], ["phonograph record", "phonograph recording", "record", "record phonograph recording"], ["piano"], ["pickle"], ["pickup truck"], ["pie"], ["pigeon"], ["pillow"], ["pineapple"], ["pinecone"], ["pipe", "piping"], ["pita", "pita bread", "pocket bread"], ["pitcher", "pitcher vessel for liquid", "ewer"], ["pizza"], ["place mat"], ["plate"], ["platter"], ["pliers", "plyers"], ["pocketknife"], ["poker", "poker fire stirring tool", "stove poker", "fire hook"], ["pole", "post"], ["polo shirt", "sport shirt"], ["pony"], ["pop", "pop soda", "soda", "soda pop", "tonic", "soft drink"], ["postbox", "postbox public", "mailbox", "mailbox public"], ["postcard", "postal card", "mailing-card"], ["poster", "placard"], ["pot"], ["flowerpot"], ["potato"], ["potholder"], ["pottery", "clayware"], ["pouch"], ["power shovel", "excavator", "digger"], ["prawn", "shrimp"], ["pretzel"], ["printer", "printing machine"], ["projectile", "projectile weapon", "missile"], ["projector"], ["propeller", "propellor"], ["pumpkin"], ["puppy"], ["quilt", "comforter"], ["rabbit"], ["racket", "racquet"], ["radiator"], ["radio receiver", "radio set", "radio", "tuner", "tuner radio"], ["radish", "daikon"], ["raft"], ["raincoat", "waterproof jacket"], ["ram", "ram animal"], ["raspberry"], ["razorblade"], ["reamer", "reamer juicer", "juicer", "juice reamer"], ["rearview mirror"], ["receipt"], ["recliner", "reclining chair", "lounger", "lounger chair"], ["record player", "phonograph", "phonograph record player", "turntable"], ["reflector"], ["remote control"], ["rhinoceros"], ["rifle"], ["ring"], ["robe"], ["rocking chair"], ["rolling pin"], ["router", "router computer equipment"], ["rubber band", "elastic band"], ["runner", "runner carpet"], ["plastic bag", "paper bag"], ["saddle", "saddle on an animal"], ["saddle blanket", "saddlecloth", "horse blanket"], ["saddlebag"], ["sail"], ["salad"], ["salami"], ["salmon", "salmon fish"], ["salsa"], ["saltshaker"], ["sandal", "sandal type of shoe"], ["sandwich"], ["saucer"], ["sausage"], ["scale", "scale measuring instrument"], ["scarf"], ["school bus"], ["scissors"], ["scoreboard"], ["screwdriver"], ["scrubbing brush"], ["sculpture"], ["seabird", "seafowl"], ["seahorse"], ["seashell"], ["sewing machine"], ["shaker"], ["shampoo"], ["shark"], ["shaving cream", "shaving soap"], ["sheep"], ["shield"], ["shirt"], ["shoe", "sneaker", "sneaker type of shoe", "tennis shoe"], ["shopping bag"], ["shopping cart"], ["short pants", "shorts", "shorts clothing", "trunks", "trunks clothing"], ["shoulder bag"], ["shovel"], ["shower head"], ["shower curtain"], ["signboard"], ["silo"], ["sink"], ["skateboard"], ["skewer"], ["ski"], ["ski boot"], ["ski parka", "ski jacket"], ["ski pole"], ["skirt"], ["sled", "sledge", "sleigh"], ["sleeping bag"], ["slipper", "slipper footwear", "carpet slipper", "carpet slipper footwear"], ["snowboard"], ["snowman"], ["snowmobile"], ["soap"], ["soccer ball"], ["sock"], ["sofa", "couch", "lounge"], ["solar array", "solar battery", "solar panel"], ["soup"], ["soupspoon"], ["sour cream", "soured cream"], ["spatula"], ["spectacles", "specs", "eyeglasses", "glasses"], ["spice rack"], ["spider"], ["sponge"], ["spoon"], ["sportswear", "athletic wear", "activewear"], ["spotlight"], ["squirrel"], ["stapler", "stapler stapling machine"], ["starfish", "sea star"], ["statue", "statue sculpture"], ["steak", "steak food"], ["steering wheel"], ["step stool"], ["stereo", "stereo sound system"], ["stirrup"], ["stool"], ["stop sign"], ["brake light"], ["stove", "kitchen stove", "range", "range kitchen appliance", "kitchen range", "cooking stove"], ["strainer"], ["strap"], ["straw", "straw for drinking", "drinking straw"], ["strawberry"], ["street sign"], ["streetlight", "street lamp"], ["suit", "suit clothing"], ["sunflower"], ["sunglasses"], ["sunhat"], ["surfboard"], ["sushi"], ["mop"], ["sweat pants"], ["sweatband"], ["sweater"], ["sweatshirt"], ["sweet potato"], ["swimsuit", "swimwear", "bathing suit", "swimming costume", "bathing costume", "swimming trunks", "bathing trunks"], ["sword"], ["table"], ["table lamp"], ["tablecloth"], ["tag"], ["taillight", "rear light"], ["tank", "tank storage vessel", "storage tank"], ["tank top", "tank top clothing"], ["tape", "tape sticky cloth or paper"], ["tape measure", "measuring tape"], ["tapestry"], ["tarp"], ["tartan", "plaid"], ["tassel"], ["tea bag"], ["teacup"], ["teakettle"], ["teapot"], ["teddy bear"], ["telephone", "phone", "telephone set"], ["telephone booth", "phone booth", "call box", "telephone box", "telephone kiosk"], ["telephone pole", "telegraph pole", "telegraph post"], ["television camera", "tv camera"], ["television set", "tv", "tv set"], ["tennis ball"], ["tennis racket"], ["thermometer"], ["thermos bottle"], ["thermostat"], ["thread", "yarn"], ["thumbtack", "drawing pin", "pushpin"], ["tiara"], ["tiger"], ["tights", "tights clothing", "leotards"], ["timer", "stopwatch"], ["tinfoil"], ["tinsel"], ["tissue paper"], ["toast", "toast food"], ["toaster"], ["toaster oven"], ["toilet"], ["toilet tissue", "toilet paper", "bathroom tissue"], ["tomato"], ["tongs"], ["toolbox"], ["toothbrush"], ["toothpaste"], ["toothpick"], ["cover"], ["tortilla"], ["tow truck"], ["towel"], ["towel rack", "towel rail", "towel bar"], ["toy"], ["tractor", "tractor farm equipment"], ["traffic light"], ["dirt bike"], ["trailer truck", "tractor trailer", "trucking rig", "articulated lorry", "semi truck"], ["train", "train railroad vehicle", "railroad train"], ["tray"], ["tricycle"], ["tripod"], ["trousers", "pants", "pants clothing"], ["truck"], ["trunk"], ["turban"], ["turkey", "turkey food"], ["turtle"], ["turtleneck", "turtleneck clothing", "polo-neck"], ["typewriter"], ["umbrella"], ["underwear", "underclothes", "underclothing", "underpants"], ["urinal"], ["urn"], ["vacuum cleaner"], ["vase"], ["vending machine"], ["vent", "blowhole", "air vent"], ["vest", "waistcoat"], ["videotape"], ["volleyball"], ["waffle"], ["wagon"], ["wagon wheel"], ["walking stick"], ["wall clock"], ["wall socket", "wall plug", "electric outlet", "electrical outlet", "outlet", "electric receptacle"], ["wallet", "billfold"], ["automatic washer", "washing machine"], ["watch", "wristwatch"], ["water bottle"], ["water cooler"], ["water faucet", "water tap", "tap", "tap water faucet"], ["water jug"], ["water scooter", "sea scooter", "jet ski"], ["water ski"], ["water tower"], ["watering can"], ["watermelon"], ["weathervane", "vane", "vane weathervane", "wind vane"], ["webcam"], ["wedding cake", "bridecake"], ["wedding ring", "wedding band"], ["wet suit"], ["wheel"], ["wheelchair"], ["whipped cream"], ["whistle"], ["wig"], ["wind chime"], ["windmill"], ["window box", "window box for plants"], ["windshield wiper", "windscreen wiper", "wiper", "wiper for windshield or screen"], ["windsock", "air sock", "air-sleeve", "wind sleeve", "wind cone"], ["wine bottle"], ["wine bucket", "wine cooler"], ["wineglass"], ["blinder", "blinder for horses"], ["wok"], ["wooden spoon"], ["wreath"], ["wrench", "spanner"], ["wristband"], ["wristlet", "wrist band"], ["yacht"], ["yogurt", "yoghurt", "yoghourt"], ["yoke", "yoke animal equipment"], ["zebra"], ["zucchini", "courgette"]]
\ No newline at end of file
data/texts/lvis_v1_class_texts.json 0 → 100644
View file @ e9cee049
[["aerosol can", "spray can"], ["air conditioner"], ["airplane", "aeroplane"], ["alarm clock"], ["alcohol", "alcoholic beverage"], ["alligator", "gator"], ["almond"], ["ambulance"], ["amplifier"], ["anklet", "ankle bracelet"], ["antenna", "aerial", "transmitting aerial"], ["apple"], ["applesauce"], ["apricot"], ["apron"], ["aquarium", "fish tank"], ["arctic", "arctic type of shoe", "galosh", "golosh", "rubber", "rubber type of shoe", "gumshoe"], ["armband"], ["armchair"], ["armoire"], ["armor", "armour"], ["artichoke"], ["trash can", "garbage can", "wastebin", "dustbin", "trash barrel", "trash bin"], ["ashtray"], ["asparagus"], ["atomizer", "atomiser", "spray", "sprayer", "nebulizer", "nebuliser"], ["avocado"], ["award", "accolade"], ["awning"], ["ax", "axe"], ["baboon"], ["baby buggy", "baby carriage", "perambulator", "pram", "stroller"], ["basketball backboard"], ["backpack", "knapsack", "packsack", "rucksack", "haversack"], ["handbag", "purse", "pocketbook"], ["suitcase", "baggage", "luggage"], ["bagel", "beigel"], ["bagpipe"], ["baguet", "baguette"], ["bait", "lure"], ["ball"], ["ballet skirt", "tutu"], ["balloon"], ["bamboo"], ["banana"], ["Band Aid"], ["bandage"], ["bandanna", "bandana"], ["banjo"], ["banner", "streamer"], ["barbell"], ["barge"], ["barrel", "cask"], ["barrette"], ["barrow", "garden cart", "lawn cart", "wheelbarrow"], ["baseball base"], ["baseball"], ["baseball bat"], ["baseball cap", "jockey cap", "golf cap"], ["baseball glove", "baseball mitt"], ["basket", "handbasket"], ["basketball"], ["bass horn", "sousaphone", "tuba"], ["bat", "bat animal"], ["bath mat"], ["bath towel"], ["bathrobe"], ["bathtub", "bathing tub"], ["batter", "batter food"], ["battery"], ["beachball"], ["bead"], ["bean curd", "tofu"], ["beanbag"], ["beanie", "beany"], ["bear"], ["bed"], ["bedpan"], ["bedspread", "bedcover", "bed covering", "counterpane", "spread"], ["cow"], ["beef", "beef food", "boeuf", "boeuf food"], ["beeper", "pager"], ["beer bottle"], ["beer can"], ["beetle"], ["bell"], ["bell pepper", "capsicum"], ["belt"], ["belt buckle"], ["bench"], ["beret"], ["bib"], ["Bible"], ["bicycle", "bike", "bike bicycle"], ["visor", "vizor"], ["billboard"], ["binder", "ring-binder"], ["binoculars", "field glasses", "opera glasses"], ["bird"], ["birdfeeder"], ["birdbath"], ["birdcage"], ["birdhouse"], ["birthday cake"], ["birthday card"], ["pirate flag"], ["black sheep"], ["blackberry"], ["blackboard", "chalkboard"], ["blanket"], ["blazer", "sport jacket", "sport coat", "sports jacket", "sports coat"], ["blender", "liquidizer", "liquidiser"], ["blimp"], ["blinker", "flasher"], ["blouse"], ["blueberry"], ["gameboard"], ["boat", "ship", "ship boat"], ["bob", "bobber", "bobfloat"], ["bobbin", "spool", "reel"], ["bobby pin", "hairgrip"], ["boiled egg", "coddled egg"], ["bolo tie", "bolo", "bola tie", "bola"], ["deadbolt"], ["bolt"], ["bonnet"], ["book"], ["bookcase"], ["booklet", "brochure", "leaflet", "pamphlet"], ["bookmark", "bookmarker"], ["boom microphone", "microphone boom"], ["boot"], ["bottle"], ["bottle opener"], ["bouquet"], ["bow", "bow weapon"], ["bow", "bow decorative ribbons"], ["bow-tie", "bowtie"], ["bowl"], ["pipe bowl"], ["bowler hat", "bowler", "derby hat", "derby", "plug hat"], ["bowling ball"], ["box"], ["boxing glove"], ["suspenders"], ["bracelet", "bangle"], ["brass plaque"], ["brassiere", "bra", "bandeau"], ["bread-bin", "breadbox"], ["bread"], ["breechcloth", "breechclout", "loincloth"], ["bridal gown", "wedding gown", "wedding dress"], ["briefcase"], ["broccoli"], ["broach"], ["broom"], ["brownie"], ["brussels sprouts"], ["bubble gum"], ["bucket", "pail"], ["horse buggy"], ["horned cow"], ["bulldog"], ["bulldozer", "dozer"], ["bullet train"], ["bulletin board", "notice board"], ["bulletproof vest"], ["bullhorn", "megaphone"], ["bun", "roll"], ["bunk bed"], ["buoy"], ["burrito"], ["bus", "bus vehicle", "autobus", "charabanc", "double-decker", "motorbus", "motorcoach"], ["business card"], ["butter"], ["butterfly"], ["button"], ["cab", "cab taxi", "taxi", "taxicab"], ["cabana"], ["cabin car", "caboose"], ["cabinet"], ["locker", "storage locker"], ["cake"], ["calculator"], ["calendar"], ["calf"], ["camcorder"], ["camel"], ["camera"], ["camera lens"], ["camper", "camper vehicle", "camping bus", "motor home"], ["can", "tin can"], ["can opener", "tin opener"], ["candle", "candlestick"], ["candle holder"], ["candy bar"], ["candy cane"], ["walking cane"], ["canister", "cannister"], ["canoe"], ["cantaloup", "cantaloupe"], ["canteen"], ["cap", "cap headwear"], ["bottle cap", "cap", "cap container lid"], ["cape"], ["cappuccino", "coffee cappuccino"], ["car", "car automobile", "auto", "auto automobile", "automobile"], ["railcar", "railcar part of a train", "railway car", "railway car part of a train", "railroad car", "railroad car part of a train"], ["elevator car"], ["car battery", "automobile battery"], ["identity card"], ["card"], ["cardigan"], ["cargo ship", "cargo vessel"], ["carnation"], ["horse carriage"], ["carrot"], ["tote bag"], ["cart"], ["carton"], ["cash register", "register", "register for cash transactions"], ["casserole"], ["cassette"], ["cast", "plaster cast", "plaster bandage"], ["cat"], ["cauliflower"], ["cayenne", "cayenne spice", "cayenne pepper", "cayenne pepper spice", "red pepper", "red pepper spice"], ["CD player"], ["celery"], ["cellular telephone", "cellular phone", "cellphone", "mobile phone", "smart phone"], ["chain mail", "ring mail", "chain armor", "chain armour", "ring armor", "ring armour"], ["chair"], ["chaise longue", "chaise", "daybed"], ["chalice"], ["chandelier"], ["chap"], ["checkbook", "chequebook"], ["checkerboard"], ["cherry"], ["chessboard"], ["chicken", "chicken animal"], ["chickpea", "garbanzo"], ["chili", "chili vegetable", "chili pepper", "chili pepper vegetable", "chilli", "chilli vegetable", "chilly", "chilly vegetable", "chile", "chile vegetable"], ["chime", "gong"], ["chinaware"], ["crisp", "crisp potato chip", "potato chip"], ["poker chip"], ["chocolate bar"], ["chocolate cake"], ["chocolate milk"], ["chocolate mousse"], ["choker", "collar", "neckband"], ["chopping board", "cutting board", "chopping block"], ["chopstick"], ["Christmas tree"], ["slide"], ["cider", "cyder"], ["cigar box"], ["cigarette"], ["cigarette case", "cigarette pack"], ["cistern", "water tank"], ["clarinet"], ["clasp"], ["cleansing agent", "cleanser", "cleaner"], ["cleat", "cleat for securing rope"], ["clementine"], ["clip"], ["clipboard"], ["clippers", "clippers for plants"], ["cloak"], ["clock", "timepiece", "timekeeper"], ["clock tower"], ["clothes hamper", "laundry basket", "clothes basket"], ["clothespin", "clothes peg"], ["clutch bag"], ["coaster"], ["coat"], ["coat hanger", "clothes hanger", "dress hanger"], ["coatrack", "hatrack"], ["cock", "rooster"], ["cockroach"], ["cocoa", "cocoa beverage", "hot chocolate", "hot chocolate beverage", "drinking chocolate"], ["coconut", "cocoanut"], ["coffee maker", "coffee machine"], ["coffee table", "cocktail table"], ["coffeepot"], ["coil"], ["coin"], ["colander", "cullender"], ["coleslaw", "slaw"], ["coloring material", "colouring material"], ["combination lock"], ["pacifier", "teething ring"], ["comic book"], ["compass"], ["computer keyboard", "keyboard", "keyboard computer"], ["condiment"], ["cone", "traffic cone"], ["control", "controller"], ["convertible", "convertible automobile"], ["sofa bed"], ["cooker"], ["cookie", "cooky", "biscuit", "biscuit cookie"], ["cooking utensil"], ["cooler", "cooler for food", "ice chest"], ["cork", "cork bottle plug", "bottle cork"], ["corkboard"], ["corkscrew", "bottle screw"], ["edible corn", "corn", "maize"], ["cornbread"], ["cornet", "horn", "trumpet"], ["cornice", "valance", "valance board", "pelmet"], ["cornmeal"], ["corset", "girdle"], ["costume"], ["cougar", "puma", "catamount", "mountain lion", "panther"], ["coverall"], ["cowbell"], ["cowboy hat", "ten-gallon hat"], ["crab", "crab animal"], ["crabmeat"], ["cracker"], ["crape", "crepe", "French pancake"], ["crate"], ["crayon", "wax crayon"], ["cream pitcher"], ["crescent roll", "croissant"], ["crib", "cot"], ["crock pot", "earthenware jar"], ["crossbar"], ["crouton"], ["crow"], ["crowbar", "wrecking bar", "pry bar"], ["crown"], ["crucifix"], ["cruise ship", "cruise liner"], ["police cruiser", "patrol car", "police car", "squad car"], ["crumb"], ["crutch"], ["cub", "cub animal"], ["cube", "square block"], ["cucumber", "cuke"], ["cufflink"], ["cup"], ["trophy cup"], ["cupboard", "closet"], ["cupcake"], ["hair curler", "hair roller", "hair crimper"], ["curling iron"], ["curtain", "drapery"], ["cushion"], ["cylinder"], ["cymbal"], ["dagger"], ["dalmatian"], ["dartboard"], ["date", "date fruit"], ["deck chair", "beach chair"], ["deer", "cervid"], ["dental floss", "floss"], ["desk"], ["detergent"], ["diaper"], ["diary", "journal"], ["die", "dice"], ["dinghy", "dory", "rowboat"], ["dining table"], ["tux", "tuxedo"], ["dish"], ["dish antenna"], ["dishrag", "dishcloth"], ["dishtowel", "tea towel"], ["dishwasher", "dishwashing machine"], ["dishwasher detergent", "dishwashing detergent", "dishwashing liquid", "dishsoap"], ["dispenser"], ["diving board"], ["Dixie cup", "paper cup"], ["dog"], ["dog collar"], ["doll"], ["dollar", "dollar bill", "one dollar bill"], ["dollhouse", "doll's house"], ["dolphin"], ["domestic ass", "donkey"], ["doorknob", "doorhandle"], ["doormat", "welcome mat"], ["doughnut", "donut"], ["dove"], ["dragonfly"], ["drawer"], ["underdrawers", "boxers", "boxershorts"], ["dress", "frock"], ["dress hat", "high hat", "opera hat", "silk hat", "top hat"], ["dress suit"], ["dresser"], ["drill"], ["drone"], ["dropper", "eye dropper"], ["drum", "drum musical instrument"], ["drumstick"], ["duck"], ["duckling"], ["duct tape"], ["duffel bag", "duffle bag", "duffel", "duffle"], ["dumbbell"], ["dumpster"], ["dustpan"], ["eagle"], ["earphone", "earpiece", "headphone"], ["earplug"], ["earring"], ["easel"], ["eclair"], ["eel"], ["egg", "eggs"], ["egg roll", "spring roll"], ["egg yolk", "yolk", "yolk egg"], ["eggbeater", "eggwhisk"], ["eggplant", "aubergine"], ["electric chair"], ["refrigerator"], ["elephant"], ["elk", "moose"], ["envelope"], ["eraser"], ["escargot"], ["eyepatch"], ["falcon"], ["fan"], ["faucet", "spigot", "tap"], ["fedora"], ["ferret"], ["Ferris wheel"], ["ferry", "ferryboat"], ["fig", "fig fruit"], ["fighter jet", "fighter aircraft", "attack aircraft"], ["figurine"], ["file cabinet", "filing cabinet"], ["file", "file tool"], ["fire alarm", "smoke alarm"], ["fire engine", "fire truck"], ["fire extinguisher", "extinguisher"], ["fire hose"], ["fireplace"], ["fireplug", "fire hydrant", "hydrant"], ["first-aid kit"], ["fish"], ["fish", "fish food"], ["fishbowl", "goldfish bowl"], ["fishing rod", "fishing pole"], ["flag"], ["flagpole", "flagstaff"], ["flamingo"], ["flannel"], ["flap"], ["flash", "flashbulb"], ["flashlight", "torch"], ["fleece"], ["flip-flop", "flip-flop sandal"], ["flipper", "flipper footwear", "fin", "fin footwear"], ["flower arrangement", "floral arrangement"], ["flute glass", "champagne flute"], ["foal"], ["folding chair"], ["food processor"], ["football", "football American"], ["football helmet"], ["footstool", "footrest"], ["fork"], ["forklift"], ["freight car"], ["French toast"], ["freshener", "air freshener"], ["frisbee"], ["frog", "toad", "toad frog"], ["fruit juice"], ["frying pan", "frypan", "skillet"], ["fudge"], ["funnel"], ["futon"], ["gag", "muzzle"], ["garbage"], ["garbage truck"], ["garden hose"], ["gargle", "mouthwash"], ["gargoyle"], ["garlic", "ail"], ["gasmask", "respirator", "gas helmet"], ["gazelle"], ["gelatin", "jelly"], ["gemstone"], ["generator"], ["giant panda", "panda", "panda bear"], ["gift wrap"], ["ginger", "gingerroot"], ["giraffe"], ["cincture", "sash", "waistband", "waistcloth"], ["glass", "glass drink container", "drinking glass"], ["globe"], ["glove"], ["goat"], ["goggles"], ["goldfish"], ["golf club", "golf-club"], ["golfcart"], ["gondola", "gondola boat"], ["goose"], ["gorilla"], ["gourd"], ["grape"], ["grater"], ["gravestone", "headstone", "tombstone"], ["gravy boat", "gravy holder"], ["green bean"], ["green onion", "spring onion", "scallion"], ["griddle"], ["grill", "grille", "grillwork", "radiator grille"], ["grits", "hominy grits"], ["grizzly", "grizzly bear"], ["grocery bag"], ["guitar"], ["gull", "seagull"], ["gun"], ["hairbrush"], ["hairnet"], ["hairpin"], ["halter top"], ["ham", "jambon", "gammon"], ["hamburger", "beefburger", "burger"], ["hammer"], ["hammock"], ["hamper"], ["hamster"], ["hair dryer"], ["hand glass", "hand mirror"], ["hand towel", "face towel"], ["handcart", "pushcart", "hand truck"], ["handcuff"], ["handkerchief"], ["handle", "grip", "handgrip"], ["handsaw", "carpenter's saw"], ["hardback book", "hardcover book"], ["harmonium", "organ", "organ musical instrument", "reed organ", "reed organ musical instrument"], ["hat"], ["hatbox"], ["veil"], ["headband"], ["headboard"], ["headlight", "headlamp"], ["headscarf"], ["headset"], ["headstall", "headstall for horses", "headpiece", "headpiece for horses"], ["heart"], ["heater", "warmer"], ["helicopter"], ["helmet"], ["heron"], ["highchair", "feeding chair"], ["hinge"], ["hippopotamus"], ["hockey stick"], ["hog", "pig"], ["home plate", "home plate baseball", "home base", "home base baseball"], ["honey"], ["fume hood", "exhaust hood"], ["hook"], ["hookah", "narghile", "nargileh", "sheesha", "shisha", "water pipe"], ["hornet"], ["horse"], ["hose", "hosepipe"], ["hot-air balloon"], ["hotplate"], ["hot sauce"], ["hourglass"], ["houseboat"], ["hummingbird"], ["hummus", "humus", "hommos", "hoummos", "humous"], ["polar bear"], ["icecream"], ["popsicle"], ["ice maker"], ["ice pack", "ice bag"], ["ice skate"], ["igniter", "ignitor", "lighter"], ["inhaler", "inhalator"], ["iPod"], ["iron", "iron for clothing", "smoothing iron", "smoothing iron for clothing"], ["ironing board"], ["jacket"], ["jam"], ["jar"], ["jean", "blue jean", "denim"], ["jeep", "landrover"], ["jelly bean", "jelly egg"], ["jersey", "T-shirt", "tee shirt"], ["jet plane", "jet-propelled plane"], ["jewel", "gem", "precious stone"], ["jewelry", "jewellery"], ["joystick"], ["jumpsuit"], ["kayak"], ["keg"], ["kennel", "doghouse"], ["kettle", "boiler"], ["key"], ["keycard"], ["kilt"], ["kimono"], ["kitchen sink"], ["kitchen table"], ["kite"], ["kitten", "kitty"], ["kiwi fruit"], ["knee pad"], ["knife"], ["knitting needle"], ["knob"], ["knocker", "knocker on a door", "doorknocker"], ["koala", "koala bear"], ["lab coat", "laboratory coat"], ["ladder"], ["ladle"], ["ladybug", "ladybeetle", "ladybird beetle"], ["lamb", "lamb animal"], ["lamb-chop", "lambchop"], ["lamp"], ["lamppost"], ["lampshade"], ["lantern"], ["lanyard", "laniard"], ["laptop computer", "notebook computer"], ["lasagna", "lasagne"], ["latch"], ["lawn mower"], ["leather"], ["legging", "legging clothing", "leging", "leging clothing", "leg covering"], ["Lego", "Lego set"], ["legume"], ["lemon"], ["lemonade"], ["lettuce"], ["license plate", "numberplate"], ["life buoy", "lifesaver", "life belt", "life ring"], ["life jacket", "life vest"], ["lightbulb"], ["lightning rod", "lightning conductor"], ["lime"], ["limousine"], ["lion"], ["lip balm"], ["liquor", "spirits", "hard liquor", "liqueur", "cordial"], ["lizard"], ["log"], ["lollipop"], ["speaker", "speaker stereo equipment"], ["loveseat"], ["machine gun"], ["magazine"], ["magnet"], ["mail slot"], ["mailbox", "mailbox at home", "letter box", "letter box at home"], ["mallard"], ["mallet"], ["mammoth"], ["manatee"], ["mandarin orange"], ["manger", "trough"], ["manhole"], ["map"], ["marker"], ["martini"], ["mascot"], ["mashed potato"], ["masher"], ["mask", "facemask"], ["mast"], ["mat", "mat gym equipment", "gym mat"], ["matchbox"], ["mattress"], ["measuring cup"], ["measuring stick", "ruler", "ruler measuring stick", "measuring rod"], ["meatball"], ["medicine"], ["melon"], ["microphone"], ["microscope"], ["microwave oven"], ["milestone", "milepost"], ["milk"], ["milk can"], ["milkshake"], ["minivan"], ["mint candy"], ["mirror"], ["mitten"], ["mixer", "mixer kitchen tool", "stand mixer"], ["money"], ["monitor", "monitor computer equipment"], ["monkey"], ["motor"], ["motor scooter", "scooter"], ["motor vehicle", "automotive vehicle"], ["motorcycle"], ["mound", "mound baseball", "pitcher's mound"], ["mouse", "mouse computer equipment", "computer mouse"], ["mousepad"], ["muffin"], ["mug"], ["mushroom"], ["music stool", "piano stool"], ["musical instrument", "instrument", "instrument musical"], ["nailfile"], ["napkin", "table napkin", "serviette"], ["neckerchief"], ["necklace"], ["necktie", "tie", "tie necktie"], ["needle"], ["nest"], ["newspaper", "paper", "paper newspaper"], ["newsstand"], ["nightshirt", "nightwear", "sleepwear", "nightclothes"], ["nosebag", "nosebag for animals", "feedbag"], ["noseband", "noseband for animals", "nosepiece", "nosepiece for animals"], ["notebook"], ["notepad"], ["nut"], ["nutcracker"], ["oar"], ["octopus", "octopus food"], ["octopus", "octopus animal"], ["oil lamp", "kerosene lamp", "kerosine lamp"], ["olive oil"], ["omelet", "omelette"], ["onion"], ["orange", "orange fruit"], ["orange juice"], ["ostrich"], ["ottoman", "pouf", "pouffe", "hassock"], ["oven"], ["overalls", "overalls clothing"], ["owl"], ["packet"], ["inkpad", "inking pad", "stamp pad"], ["pad"], ["paddle", "boat paddle"], ["padlock"], ["paintbrush"], ["painting"], ["pajamas", "pyjamas"], ["palette", "pallet"], ["pan", "pan for cooking", "cooking pan"], ["pan", "pan metal container"], ["pancake"], ["pantyhose"], ["papaya"], ["paper plate"], ["paper towel"], ["paperback book", "paper-back book", "softback book", "soft-cover book"], ["paperweight"], ["parachute"], ["parakeet", "parrakeet", "parroket", "paraquet", "paroquet", "parroquet"], ["parasail", "parasail sports"], ["parasol", "sunshade"], ["parchment"], ["parka", "anorak"], ["parking meter"], ["parrot"], ["passenger car", "passenger car part of a train", "coach", "coach part of a train"], ["passenger ship"], ["passport"], ["pastry"], ["patty", "patty food"], ["pea", "pea food"], ["peach"], ["peanut butter"], ["pear"], ["peeler", "peeler tool for fruit and vegetables"], ["wooden leg", "pegleg"], ["pegboard"], ["pelican"], ["pen"], ["pencil"], ["pencil box", "pencil case"], ["pencil sharpener"], ["pendulum"], ["penguin"], ["pennant"], ["penny", "penny coin"], ["pepper", "peppercorn"], ["pepper mill", "pepper grinder"], ["perfume"], ["persimmon"], ["person", "baby", "child", "boy", "girl", "man", "woman", "human"], ["pet"], ["pew", "pew church bench", "church bench"], ["phonebook", "telephone book", "telephone directory"], ["phonograph record", "phonograph recording", "record", "record phonograph recording"], ["piano"], ["pickle"], ["pickup truck"], ["pie"], ["pigeon"], ["piggy bank", "penny bank"], ["pillow"], ["pin", "pin non jewelry"], ["pineapple"], ["pinecone"], ["ping-pong ball"], ["pinwheel"], ["tobacco pipe"], ["pipe", "piping"], ["pistol", "handgun"], ["pita", "pita bread", "pocket bread"], ["pitcher", "pitcher vessel for liquid", "ewer"], ["pitchfork"], ["pizza"], ["place mat"], ["plate"], ["platter"], ["playpen"], ["pliers", "plyers"], ["plow", "plow farm equipment", "plough", "plough farm equipment"], ["plume"], ["pocket watch"], ["pocketknife"], ["poker", "poker fire stirring tool", "stove poker", "fire hook"], ["pole", "post"], ["polo shirt", "sport shirt"], ["poncho"], ["pony"], ["pool table", "billiard table", "snooker table"], ["pop", "pop soda", "soda", "soda pop", "tonic", "soft drink"], ["postbox", "postbox public", "mailbox", "mailbox public"], ["postcard", "postal card", "mailing-card"], ["poster", "placard"], ["pot"], ["flowerpot"], ["potato"], ["potholder"], ["pottery", "clayware"], ["pouch"], ["power shovel", "excavator", "digger"], ["prawn", "shrimp"], ["pretzel"], ["printer", "printing machine"], ["projectile", "projectile weapon", "missile"], ["projector"], ["propeller", "propellor"], ["prune"], ["pudding"], ["puffer", "puffer fish", "pufferfish", "blowfish", "globefish"], ["puffin"], ["pug-dog"], ["pumpkin"], ["puncher"], ["puppet", "marionette"], ["puppy"], ["quesadilla"], ["quiche"], ["quilt", "comforter"], ["rabbit"], ["race car", "racing car"], ["racket", "racquet"], ["radar"], ["radiator"], ["radio receiver", "radio set", "radio", "tuner", "tuner radio"], ["radish", "daikon"], ["raft"], ["rag doll"], ["raincoat", "waterproof jacket"], ["ram", "ram animal"], ["raspberry"], ["rat"], ["razorblade"], ["reamer", "reamer juicer", "juicer", "juice reamer"], ["rearview mirror"], ["receipt"], ["recliner", "reclining chair", "lounger", "lounger chair"], ["record player", "phonograph", "phonograph record player", "turntable"], ["reflector"], ["remote control"], ["rhinoceros"], ["rib", "rib food"], ["rifle"], ["ring"], ["river boat"], ["road map"], ["robe"], ["rocking chair"], ["rodent"], ["roller skate"], ["Rollerblade"], ["rolling pin"], ["root beer"], ["router", "router computer equipment"], ["rubber band", "elastic band"], ["runner", "runner carpet"], ["plastic bag", "paper bag"], ["saddle", "saddle on an animal"], ["saddle blanket", "saddlecloth", "horse blanket"], ["saddlebag"], ["safety pin"], ["sail"], ["salad"], ["salad plate", "salad bowl"], ["salami"], ["salmon", "salmon fish"], ["salmon", "salmon food"], ["salsa"], ["saltshaker"], ["sandal", "sandal type of shoe"], ["sandwich"], ["satchel"], ["saucepan"], ["saucer"], ["sausage"], ["sawhorse", "sawbuck"], ["saxophone"], ["scale", "scale measuring instrument"], ["scarecrow", "strawman"], ["scarf"], ["school bus"], ["scissors"], ["scoreboard"], ["scraper"], ["screwdriver"], ["scrubbing brush"], ["sculpture"], ["seabird", "seafowl"], ["seahorse"], ["seaplane", "hydroplane"], ["seashell"], ["sewing machine"], ["shaker"], ["shampoo"], ["shark"], ["sharpener"], ["Sharpie"], ["shaver", "shaver electric", "electric shaver", "electric razor"], ["shaving cream", "shaving soap"], ["shawl"], ["shears"], ["sheep"], ["shepherd dog", "sheepdog"], ["sherbert", "sherbet"], ["shield"], ["shirt"], ["shoe", "sneaker", "sneaker type of shoe", "tennis shoe"], ["shopping bag"], ["shopping cart"], ["short pants", "shorts", "shorts clothing", "trunks", "trunks clothing"], ["shot glass"], ["shoulder bag"], ["shovel"], ["shower head"], ["shower cap"], ["shower curtain"], ["shredder", "shredder for paper"], ["signboard"], ["silo"], ["sink"], ["skateboard"], ["skewer"], ["ski"], ["ski boot"], ["ski parka", "ski jacket"], ["ski pole"], ["skirt"], ["skullcap"], ["sled", "sledge", "sleigh"], ["sleeping bag"], ["sling", "sling bandage", "triangular bandage"], ["slipper", "slipper footwear", "carpet slipper", "carpet slipper footwear"], ["smoothie"], ["snake", "serpent"], ["snowboard"], ["snowman"], ["snowmobile"], ["soap"], ["soccer ball"], ["sock"], ["sofa", "couch", "lounge"], ["softball"], ["solar array", "solar battery", "solar panel"], ["sombrero"], ["soup"], ["soup bowl"], ["soupspoon"], ["sour cream", "soured cream"], ["soya milk", "soybean milk", "soymilk"], ["space shuttle"], ["sparkler", "sparkler fireworks"], ["spatula"], ["spear", "lance"], ["spectacles", "specs", "eyeglasses", "glasses"], ["spice rack"], ["spider"], ["crawfish", "crayfish"], ["sponge"], ["spoon"], ["sportswear", "athletic wear", "activewear"], ["spotlight"], ["squid", "squid food", "calamari", "calamary"], ["squirrel"], ["stagecoach"], ["stapler", "stapler stapling machine"], ["starfish", "sea star"], ["statue", "statue sculpture"], ["steak", "steak food"], ["steak knife"], ["steering wheel"], ["stepladder"], ["step stool"], ["stereo", "stereo sound system"], ["stew"], ["stirrer"], ["stirrup"], ["stool"], ["stop sign"], ["brake light"], ["stove", "kitchen stove", "range", "range kitchen appliance", "kitchen range", "cooking stove"], ["strainer"], ["strap"], ["straw", "straw for drinking", "drinking straw"], ["strawberry"], ["street sign"], ["streetlight", "street lamp"], ["string cheese"], ["stylus"], ["subwoofer"], ["sugar bowl"], ["sugarcane", "sugarcane plant"], ["suit", "suit clothing"], ["sunflower"], ["sunglasses"], ["sunhat"], ["surfboard"], ["sushi"], ["mop"], ["sweat pants"], ["sweatband"], ["sweater"], ["sweatshirt"], ["sweet potato"], ["swimsuit", "swimwear", "bathing suit", "swimming costume", "bathing costume", "swimming trunks", "bathing trunks"], ["sword"], ["syringe"], ["Tabasco sauce"], ["table-tennis table", "ping-pong table"], ["table"], ["table lamp"], ["tablecloth"], ["tachometer"], ["taco"], ["tag"], ["taillight", "rear light"], ["tambourine"], ["army tank", "armored combat vehicle", "armoured combat vehicle"], ["tank", "tank storage vessel", "storage tank"], ["tank top", "tank top clothing"], ["tape", "tape sticky cloth or paper"], ["tape measure", "measuring tape"], ["tapestry"], ["tarp"], ["tartan", "plaid"], ["tassel"], ["tea bag"], ["teacup"], ["teakettle"], ["teapot"], ["teddy bear"], ["telephone", "phone", "telephone set"], ["telephone booth", "phone booth", "call box", "telephone box", "telephone kiosk"], ["telephone pole", "telegraph pole", "telegraph post"], ["telephoto lens", "zoom lens"], ["television camera", "tv camera"], ["television set", "tv", "tv set"], ["tennis ball"], ["tennis racket"], ["tequila"], ["thermometer"], ["thermos bottle"], ["thermostat"], ["thimble"], ["thread", "yarn"], ["thumbtack", "drawing pin", "pushpin"], ["tiara"], ["tiger"], ["tights", "tights clothing", "leotards"], ["timer", "stopwatch"], ["tinfoil"], ["tinsel"], ["tissue paper"], ["toast", "toast food"], ["toaster"], ["toaster oven"], ["toilet"], ["toilet tissue", "toilet paper", "bathroom tissue"], ["tomato"], ["tongs"], ["toolbox"], ["toothbrush"], ["toothpaste"], ["toothpick"], ["cover"], ["tortilla"], ["tow truck"], ["towel"], ["towel rack", "towel rail", "towel bar"], ["toy"], ["tractor", "tractor farm equipment"], ["traffic light"], ["dirt bike"], ["trailer truck", "tractor trailer", "trucking rig", "articulated lorry", "semi truck"], ["train", "train railroad vehicle", "railroad train"], ["trampoline"], ["tray"], ["trench coat"], ["triangle", "triangle musical instrument"], ["tricycle"], ["tripod"], ["trousers", "pants", "pants clothing"], ["truck"], ["truffle", "truffle chocolate", "chocolate truffle"], ["trunk"], ["vat"], ["turban"], ["turkey", "turkey food"], ["turnip"], ["turtle"], ["turtleneck", "turtleneck clothing", "polo-neck"], ["typewriter"], ["umbrella"], ["underwear", "underclothes", "underclothing", "underpants"], ["unicycle"], ["urinal"], ["urn"], ["vacuum cleaner"], ["vase"], ["vending machine"], ["vent", "blowhole", "air vent"], ["vest", "waistcoat"], ["videotape"], ["vinegar"], ["violin", "fiddle"], ["vodka"], ["volleyball"], ["vulture"], ["waffle"], ["waffle iron"], ["wagon"], ["wagon wheel"], ["walking stick"], ["wall clock"], ["wall socket", "wall plug", "electric outlet", "electrical outlet", "outlet", "electric receptacle"], ["wallet", "billfold"], ["walrus"], ["wardrobe"], ["washbasin", "basin", "basin for washing", "washbowl", "washstand", "handbasin"], ["automatic washer", "washing machine"], ["watch", "wristwatch"], ["water bottle"], ["water cooler"], ["water faucet", "water tap", "tap", "tap water faucet"], ["water heater", "hot-water heater"], ["water jug"], ["water gun", "squirt gun"], ["water scooter", "sea scooter", "jet ski"], ["water ski"], ["water tower"], ["watering can"], ["watermelon"], ["weathervane", "vane", "vane weathervane", "wind vane"], ["webcam"], ["wedding cake", "bridecake"], ["wedding ring", "wedding band"], ["wet suit"], ["wheel"], ["wheelchair"], ["whipped cream"], ["whistle"], ["wig"], ["wind chime"], ["windmill"], ["window box", "window box for plants"], ["windshield wiper", "windscreen wiper", "wiper", "wiper for windshield or screen"], ["windsock", "air sock", "air-sleeve", "wind sleeve", "wind cone"], ["wine bottle"], ["wine bucket", "wine cooler"], ["wineglass"], ["blinder", "blinder for horses"], ["wok"], ["wolf"], ["wooden spoon"], ["wreath"], ["wrench", "spanner"], ["wristband"], ["wristlet", "wrist band"], ["yacht"], ["yogurt", "yoghurt", "yoghourt"], ["yoke", "yoke animal equipment"], ["zebra"], ["zucchini", "courgette"]]
\ No newline at end of file
data/texts/obj365v1_class_texts.json 0 → 100644
View file @ e9cee049
[["person"], ["sneakers"], ["chair"], ["hat"], ["lamp"], ["bottle"], ["cabinet", "shelf"], ["cup"], ["car"], ["glasses"], ["picture", "frame"], ["desk"], ["handbag"], ["street lights"], ["book"], ["plate"], ["helmet"], ["leather shoes"], ["pillow"], ["glove"], ["potted plant"], ["bracelet"], ["flower"], ["tv"], ["storage box"], ["vase"], ["bench"], ["wine glass"], ["boots"], ["bowl"], ["dining table"], ["umbrella"], ["boat"], ["flag"], ["speaker"], ["trash bin", "can"], ["stool"], ["backpack"], ["couch"], ["belt"], ["carpet"], ["basket"], ["towel", "napkin"], ["slippers"], ["barrel", "bucket"], ["coffee table"], ["suv"], ["toy"], ["tie"], ["bed"], ["traffic light"], ["pen", "pencil"], ["microphone"], ["sandals"], ["canned"], ["necklace"], ["mirror"], ["faucet"], ["bicycle"], ["bread"], ["high heels"], ["ring"], ["van"], ["watch"], ["sink"], ["horse"], ["fish"], ["apple"], ["camera"], ["candle"], ["teddy bear"], ["cake"], ["motorcycle"], ["wild bird"], ["laptop"], ["knife"], ["traffic sign"], ["cell phone"], ["paddle"], ["truck"], ["cow"], ["power outlet"], ["clock"], ["drum"], ["fork"], ["bus"], ["hanger"], ["nightstand"], ["pot", "pan"], ["sheep"], ["guitar"], ["traffic cone"], ["tea pot"], ["keyboard"], ["tripod"], ["hockey"], ["fan"], ["dog"], ["spoon"], ["blackboard", "whiteboard"], ["balloon"], ["air conditioner"], ["cymbal"], ["mouse"], ["telephone"], ["pickup truck"], ["orange"], ["banana"], ["airplane"], ["luggage"], ["skis"], ["soccer"], ["trolley"], ["oven"], ["remote"], ["baseball glove"], ["paper towel"], ["refrigerator"], ["train"], ["tomato"], ["machinery vehicle"], ["tent"], ["shampoo", "shower gel"], ["head phone"], ["lantern"], ["donut"], ["cleaning products"], ["sailboat"], ["tangerine"], ["pizza"], ["kite"], ["computer box"], ["elephant"], ["toiletries"], ["gas stove"], ["broccoli"], ["toilet"], ["stroller"], ["shovel"], ["baseball bat"], ["microwave"], ["skateboard"], ["surfboard"], ["surveillance camera"], ["gun"], ["life saver"], ["cat"], ["lemon"], ["liquid soap"], ["zebra"], ["duck"], ["sports car"], ["giraffe"], ["pumpkin"], ["piano"], ["stop sign"], ["radiator"], ["converter"], ["tissue"], ["carrot"], ["washing machine"], ["vent"], ["cookies"], ["cutting", "chopping board"], ["tennis racket"], ["candy"], ["skating and skiing shoes"], ["scissors"], ["folder"], ["baseball"], ["strawberry"], ["bow tie"], ["pigeon"], ["pepper"], ["coffee machine"], ["bathtub"], ["snowboard"], ["suitcase"], ["grapes"], ["ladder"], ["pear"], ["american football"], ["basketball"], ["potato"], ["paint brush"], ["printer"], ["billiards"], ["fire hydrant"], ["goose"], ["projector"], ["sausage"], ["fire extinguisher"], ["extension cord"], ["facial mask"], ["tennis ball"], ["chopsticks"], ["electronic stove and gas stove"], ["pie"], ["frisbee"], ["kettle"], ["hamburger"], ["golf club"], ["cucumber"], ["clutch"], ["blender"], ["tong"], ["slide"], ["hot dog"], ["toothbrush"], ["facial cleanser"], ["mango"], ["deer"], ["egg"], ["violin"], ["marker"], ["ship"], ["chicken"], ["onion"], ["ice cream"], ["tape"], ["wheelchair"], ["plum"], ["bar soap"], ["scale"], ["watermelon"], ["cabbage"], ["router", "modem"], ["golf ball"], ["pine apple"], ["crane"], ["fire truck"], ["peach"], ["cello"], ["notepaper"], ["tricycle"], ["toaster"], ["helicopter"], ["green beans"], ["brush"], ["carriage"], ["cigar"], ["earphone"], ["penguin"], ["hurdle"], ["swing"], ["radio"], ["cd"], ["parking meter"], ["swan"], ["garlic"], ["french fries"], ["horn"], ["avocado"], ["saxophone"], ["trumpet"], ["sandwich"], ["cue"], ["kiwi fruit"], ["bear"], ["fishing rod"], ["cherry"], ["tablet"], ["green vegetables"], ["nuts"], ["corn"], ["key"], ["screwdriver"], ["globe"], ["broom"], ["pliers"], ["volleyball"], ["hammer"], ["eggplant"], ["trophy"], ["dates"], ["board eraser"], ["rice"], ["tape measure", "ruler"], ["dumbbell"], ["hamimelon"], ["stapler"], ["camel"], ["lettuce"], ["goldfish"], ["meat balls"], ["medal"], ["toothpaste"], ["antelope"], ["shrimp"], ["rickshaw"], ["trombone"], ["pomegranate"], ["coconut"], ["jellyfish"], ["mushroom"], ["calculator"], ["treadmill"], ["butterfly"], ["egg tart"], ["cheese"], ["pig"], ["pomelo"], ["race car"], ["rice cooker"], ["tuba"], ["crosswalk sign"], ["papaya"], ["hair drier"], ["green onion"], ["chips"], ["dolphin"], ["sushi"], ["urinal"], ["donkey"], ["electric drill"], ["spring rolls"], ["tortoise", "turtle"], ["parrot"], ["flute"], ["measuring cup"], ["shark"], ["steak"], ["poker card"], ["binoculars"], ["llama"], ["radish"], ["noodles"], ["yak"], ["mop"], ["crab"], ["microscope"], ["barbell"], ["bread", "bun"], ["baozi"], ["lion"], ["red cabbage"], ["polar bear"], ["lighter"], ["seal"], ["mangosteen"], ["comb"], ["eraser"], ["pitaya"], ["scallop"], ["pencil case"], ["saw"], ["table tennis paddle"], ["okra"], ["starfish"], ["eagle"], ["monkey"], ["durian"], ["game board"], ["rabbit"], ["french horn"], ["ambulance"], ["asparagus"], ["hoverboard"], ["pasta"], ["target"], ["hotair balloon"], ["chainsaw"], ["lobster"], ["iron"], ["flashlight"]]
\ No newline at end of file
demo/README.md 0 → 100644
View file @ e9cee049
## YOLO-World Demo
### Getting Started
Setting `PYTHONPATH` as the path to `YOLO-World` and run:
```bash
PYTHONPATH=/xxxx/YOLO-World python demo/yyyy_demo.py
# or directly
PYTHONPATH=./ python demo/yyyy_demo.py
```
#### Gradio Demo
We provide the [Gradio](https://www.gradio.app/) demo for local devices:
```bash
pip install gradio==4.16.0
python demo/demo.py path/to/config path/to/weights
```
Additionaly, you can use a Dockerfile to build an image with gradio. As a prerequisite, make sure you have respective drivers installed alongside [nvidia-container-runtime](https://stackoverflow.com/questions/59691207/docker-build-with-nvidia-runtime). Replace MODEL_NAME and WEIGHT_NAME with the respective values or ommit this and use default values from the [Dockerfile](Dockerfile#3)
```bash
docker build --build-arg="MODEL=MODEL_NAME" --build-arg="WEIGHT=WEIGHT_NAME" -t yolo_demo .
docker run --runtime nvidia -p 8080:8080
```
#### Image Demo
We provide a simple image demo for inference on images with visualization outputs.
```bash
python demo/image_demo.py path/to/config path/to/weights image/path/directory 'person,dog,cat' --topk 100 --threshold 0.005 --output-dir demo_outputs
```
**Notes:**
* The `image` can be a directory or a single image.
* The `texts` can be a string of categories (noun phrases) which is separated by a comma. We also support `txt` file in which each line contains a category ( noun phrases).
* The `topk` and `threshold` control the number of predictions and the confidence threshold.
#### Video Demo
The `video_demo` has similar hyper-parameters with `image_demo`.
```bash
python demo/video_demo.py path/to/config path/to/weights video_path 'person,dog' --out out_video_path
```
### FAQ
> 1. `Failed to custom import!`
```bash
File "simple_demo.py", line 37, in <module>
cfg = Config.fromfile(config_file)
File "/data/miniconda3/envs/det/lib/python3.8/site-packages/mmengine/config/config.py", line 183, in fromfile
raise ImportError('Failed to custom import!') from e
ImportError: Failed to custom import!
```
**Solution:**
```bash
PYTHONPATH=/xxxx/YOLO-World python demo/simple_demo.py
```
\ No newline at end of file
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment