Skip to content

GitLab

Commit 5b3e36dc authored by Sugon_ldc's avatar Sugon_ldc
Browse files

add model TSM

parents
Pipeline #315 failed with stages
in 0 seconds
Hide whitespace changes
Inline Side-by-side
Showing with 2345 additions and 0 deletions
configs/recognition/slowfast/slowfast_prebn_r50_8x8x1_256e_kinetics400_rgb_steplr.py 0 → 100644
View file @ 5b3e36dc
_base_ = ['./slowfast_r50_8x8x1_256e_kinetics400_rgb.py']
model = dict(backbone=dict(slow_pathway=dict(lateral_norm=True)))
lr_config = dict(
policy='step',
min_lr=0,
warmup='linear',
warmup_by_epoch=True,
warmup_iters=34,
step=[94, 154, 196])
precise_bn = dict(num_iters=200, interval=5)
work_dir = './work_dirs/slowfast_prebn_r50_8x8x1_256e_kinetics400_rgb_steplr'
configs/recognition/slowfast/slowfast_r101_8x8x1_256e_kinetics400_rgb.py 0 → 100644
View file @ 5b3e36dc
model = dict(
type='Recognizer3D',
backbone=dict(
type='ResNet3dSlowFast',
pretrained=None,
resample_rate=4, # tau
speed_ratio=4, # alpha
channel_ratio=8, # beta_inv
slow_pathway=dict(
type='resnet3d',
depth=101,
pretrained=None,
lateral=True,
fusion_kernel=7,
conv1_kernel=(1, 7, 7),
dilations=(1, 1, 1, 1),
conv1_stride_t=1,
pool1_stride_t=1,
inflate=(0, 0, 1, 1),
norm_eval=False),
fast_pathway=dict(
type='resnet3d',
depth=101,
pretrained=None,
lateral=False,
base_channels=8,
conv1_kernel=(5, 7, 7),
conv1_stride_t=1,
pool1_stride_t=1,
norm_eval=False)),
cls_head=dict(
type='SlowFastHead',
in_channels=2304, # 2048+256
num_classes=400,
spatial_type='avg',
dropout_ratio=0.5),
train_cfg=None,
test_cfg=dict(average_clips='prob', max_testing_views=10))
dataset_type = 'RawframeDataset'
data_root = 'data/kinetics400/rawframes_train'
data_root_val = 'data/kinetics400/rawframes_val'
ann_file_train = 'data/kinetics400/kinetics400_train_list_rawframes.txt'
ann_file_val = 'data/kinetics400/kinetics400_val_list_rawframes.txt'
ann_file_test = 'data/kinetics400/kinetics400_val_list_rawframes.txt'
img_norm_cfg = dict(
mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_bgr=False)
train_pipeline = [
dict(type='SampleFrames', clip_len=32, frame_interval=2, num_clips=1),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='RandomResizedCrop'),
dict(type='Resize', scale=(224, 224), keep_ratio=False),
dict(type='Flip', flip_ratio=0.5),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs', 'label'])
]
val_pipeline = [
dict(
type='SampleFrames',
clip_len=32,
frame_interval=2,
num_clips=1,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='CenterCrop', crop_size=224),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
test_pipeline = [
dict(
type='SampleFrames',
clip_len=32,
frame_interval=2,
num_clips=10,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='ThreeCrop', crop_size=256),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
data = dict(
videos_per_gpu=8,
workers_per_gpu=2,
test_dataloader=dict(videos_per_gpu=1),
train=dict(
type=dataset_type,
ann_file=ann_file_train,
data_prefix=data_root,
pipeline=train_pipeline),
val=dict(
type=dataset_type,
ann_file=ann_file_val,
data_prefix=data_root_val,
pipeline=val_pipeline),
test=dict(
type=dataset_type,
ann_file=ann_file_test,
data_prefix=data_root_val,
pipeline=test_pipeline))
# optimizer
optimizer = dict(
type='SGD', lr=0.1, momentum=0.9,
weight_decay=0.0001) # this lr is used for 8 gpus
optimizer_config = dict(grad_clip=dict(max_norm=40, norm_type=2))
# learning policy
lr_config = dict(
policy='CosineAnnealing',
min_lr=0,
warmup='linear',
warmup_by_epoch=True,
warmup_iters=34)
total_epochs = 256
checkpoint_config = dict(interval=4)
workflow = [('train', 1)]
evaluation = dict(
interval=5, metrics=['top_k_accuracy', 'mean_class_accuracy'])
log_config = dict(
interval=20,
hooks=[
dict(type='TextLoggerHook'),
# dict(type='TensorboardLoggerHook'),
])
dist_params = dict(backend='nccl')
log_level = 'INFO'
work_dir = './work_dirs/slowfast_r101_8x8x1_256e_kinetics400_rgb'
load_from = None
resume_from = None
find_unused_parameters = False
configs/recognition/slowfast/slowfast_r101_r50_4x16x1_256e_kinetics400_rgb.py 0 → 100644
View file @ 5b3e36dc
model = dict(
type='Recognizer3D',
backbone=dict(
type='ResNet3dSlowFast',
pretrained=None,
resample_rate=8, # tau
speed_ratio=8, # alpha
channel_ratio=8, # beta_inv
slow_pathway=dict(
type='resnet3d',
depth=101,
pretrained=None,
lateral=True,
conv1_kernel=(1, 7, 7),
dilations=(1, 1, 1, 1),
conv1_stride_t=1,
pool1_stride_t=1,
inflate=(0, 0, 1, 1),
norm_eval=False),
fast_pathway=dict(
type='resnet3d',
depth=50,
pretrained=None,
lateral=False,
base_channels=8,
conv1_kernel=(5, 7, 7),
conv1_stride_t=1,
pool1_stride_t=1,
norm_eval=False)),
cls_head=dict(
type='SlowFastHead',
in_channels=2304, # 2048+256
num_classes=400,
spatial_type='avg',
dropout_ratio=0.5),
train_cfg=None,
test_cfg=dict(average_clips='prob', max_testing_views=10))
dataset_type = 'RawframeDataset'
data_root = 'data/kinetics400/rawframes_train'
data_root_val = 'data/kinetics400/rawframes_val'
ann_file_train = 'data/kinetics400/kinetics400_train_list_rawframes.txt'
ann_file_val = 'data/kinetics400/kinetics400_val_list_rawframes.txt'
ann_file_test = 'data/kinetics400/kinetics400_val_list_rawframes.txt'
img_norm_cfg = dict(
mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_bgr=False)
train_pipeline = [
dict(type='SampleFrames', clip_len=32, frame_interval=2, num_clips=1),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='RandomResizedCrop'),
dict(type='Resize', scale=(224, 224), keep_ratio=False),
dict(type='Flip', flip_ratio=0.5),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs', 'label'])
]
val_pipeline = [
dict(
type='SampleFrames',
clip_len=32,
frame_interval=2,
num_clips=1,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='CenterCrop', crop_size=224),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
test_pipeline = [
dict(
type='SampleFrames',
clip_len=32,
frame_interval=2,
num_clips=10,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='ThreeCrop', crop_size=256),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
data = dict(
videos_per_gpu=8,
workers_per_gpu=2,
test_dataloader=dict(videos_per_gpu=1),
train=dict(
type=dataset_type,
ann_file=ann_file_train,
data_prefix=data_root,
pipeline=train_pipeline),
val=dict(
type=dataset_type,
ann_file=ann_file_val,
data_prefix=data_root_val,
pipeline=val_pipeline),
test=dict(
type=dataset_type,
ann_file=ann_file_test,
data_prefix=data_root_val,
pipeline=test_pipeline))
# optimizer
optimizer = dict(
type='SGD', lr=0.1, momentum=0.9,
weight_decay=0.0001) # this lr is used for 8 gpus
optimizer_config = dict(grad_clip=dict(max_norm=40, norm_type=2))
# learning policy
lr_config = dict(
policy='CosineAnnealing',
min_lr=0,
warmup='linear',
warmup_by_epoch=True,
warmup_iters=34)
total_epochs = 256
checkpoint_config = dict(interval=4)
workflow = [('train', 1)]
evaluation = dict(
interval=5, metrics=['top_k_accuracy', 'mean_class_accuracy'])
log_config = dict(
interval=20,
hooks=[
dict(type='TextLoggerHook'),
# dict(type='TensorboardLoggerHook'),
])
dist_params = dict(backend='nccl')
log_level = 'INFO'
work_dir = './work_dirs/slowfast_r101_r50_4x16x1_256e_kinetics400_rgb'
load_from = None
resume_from = None
find_unused_parameters = False
configs/recognition/slowfast/slowfast_r152_r50_4x16x1_256e_kinetics400_rgb.py 0 → 100644
View file @ 5b3e36dc
model = dict(
type='Recognizer3D',
backbone=dict(
type='ResNet3dSlowFast',
pretrained=None,
resample_rate=8, # tau
speed_ratio=8, # alpha
channel_ratio=8, # beta_inv
slow_pathway=dict(
type='resnet3d',
depth=152,
pretrained=None,
lateral=True,
conv1_kernel=(1, 7, 7),
dilations=(1, 1, 1, 1),
conv1_stride_t=1,
pool1_stride_t=1,
inflate=(0, 0, 1, 1),
norm_eval=False),
fast_pathway=dict(
type='resnet3d',
depth=50,
pretrained=None,
lateral=False,
base_channels=8,
conv1_kernel=(5, 7, 7),
conv1_stride_t=1,
pool1_stride_t=1,
norm_eval=False)),
cls_head=dict(
type='SlowFastHead',
in_channels=2304, # 2048+256
num_classes=400,
spatial_type='avg',
dropout_ratio=0.5),
train_cfg=None,
test_cfg=dict(average_clips='prob', max_testing_views=8))
dataset_type = 'RawframeDataset'
data_root = 'data/kinetics400/rawframes_train'
data_root_val = 'data/kinetics400/rawframes_val'
ann_file_train = 'data/kinetics400/kinetics400_train_list_rawframes.txt'
ann_file_val = 'data/kinetics400/kinetics400_val_list_rawframes.txt'
ann_file_test = 'data/kinetics400/kinetics400_val_list_rawframes.txt'
img_norm_cfg = dict(
mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_bgr=False)
train_pipeline = [
dict(type='SampleFrames', clip_len=32, frame_interval=2, num_clips=1),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='RandomResizedCrop'),
dict(type='Resize', scale=(224, 224), keep_ratio=False),
dict(type='Flip', flip_ratio=0.5),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs', 'label'])
]
val_pipeline = [
dict(
type='SampleFrames',
clip_len=32,
frame_interval=2,
num_clips=1,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='CenterCrop', crop_size=224),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
test_pipeline = [
dict(
type='SampleFrames',
clip_len=32,
frame_interval=2,
num_clips=10,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='ThreeCrop', crop_size=256),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
data = dict(
videos_per_gpu=8,
workers_per_gpu=2,
test_dataloader=dict(videos_per_gpu=1),
train=dict(
type=dataset_type,
ann_file=ann_file_train,
data_prefix=data_root,
pipeline=train_pipeline),
val=dict(
type=dataset_type,
ann_file=ann_file_val,
data_prefix=data_root_val,
pipeline=val_pipeline),
test=dict(
type=dataset_type,
ann_file=ann_file_test,
data_prefix=data_root_val,
pipeline=test_pipeline))
# optimizer
optimizer = dict(
type='SGD', lr=0.1, momentum=0.9,
weight_decay=0.0001) # this lr is used for 8 gpus
optimizer_config = dict(grad_clip=dict(max_norm=40, norm_type=2))
# learning policy
lr_config = dict(
policy='CosineAnnealing',
min_lr=0,
warmup='linear',
warmup_by_epoch=True,
warmup_iters=34)
total_epochs = 256
checkpoint_config = dict(interval=4)
workflow = [('train', 1)]
evaluation = dict(
interval=5, metrics=['top_k_accuracy', 'mean_class_accuracy'])
log_config = dict(
interval=20,
hooks=[
dict(type='TextLoggerHook'),
# dict(type='TensorboardLoggerHook'),
])
dist_params = dict(backend='nccl')
log_level = 'INFO'
work_dir = './work_dirs/slowfast_r152_r50_4x16x1_256e_kinetics400_rgb'
load_from = None
resume_from = None
find_unused_parameters = False
configs/recognition/slowfast/slowfast_r50_16x8x1_22e_sthv1_rgb.py 0 → 100644
View file @ 5b3e36dc
_base_ = [
'../../_base_/models/slowfast_r50.py', '../../_base_/default_runtime.py'
]
model = dict(
backbone=dict(
resample_rate=4, # tau
speed_ratio=4, # alpha
channel_ratio=8, # beta_inv
slow_pathway=dict(fusion_kernel=7)),
cls_head=dict(num_classes=174))
# dataset settings
dataset_type = 'RawframeDataset'
data_root = 'data/sthv1/rawframes'
data_root_val = 'data/sthv1/rawframes'
ann_file_train = 'data/sthv1/sthv1_train_list_rawframes.txt'
ann_file_val = 'data/sthv1/sthv1_val_list_rawframes.txt'
ann_file_test = 'data/sthv1/sthv1_val_list_rawframes.txt'
sthv1_flip_label_map = {2: 4, 4: 2, 30: 41, 41: 30, 52: 66, 66: 52}
img_norm_cfg = dict(
mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_bgr=False)
train_pipeline = [
dict(type='SampleFrames', clip_len=64, frame_interval=2, num_clips=1),
dict(type='RawFrameDecode'),
dict(type='RandomResizedCrop'),
dict(type='Resize', scale=(224, 224), keep_ratio=False),
dict(type='Flip', flip_ratio=0.5, flip_label_map=sthv1_flip_label_map),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs', 'label'])
]
val_pipeline = [
dict(
type='SampleFrames',
clip_len=64,
frame_interval=2,
num_clips=1,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='CenterCrop', crop_size=256),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
test_pipeline = [
dict(
type='SampleFrames',
clip_len=64,
frame_interval=2,
num_clips=1,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='CenterCrop', crop_size=256),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
data = dict(
videos_per_gpu=4,
workers_per_gpu=2,
test_dataloader=dict(videos_per_gpu=1),
train=dict(
type=dataset_type,
ann_file=ann_file_train,
data_prefix=data_root,
filename_tmpl='{:05}.jpg',
pipeline=train_pipeline),
val=dict(
type=dataset_type,
ann_file=ann_file_val,
data_prefix=data_root_val,
filename_tmpl='{:05}.jpg',
pipeline=val_pipeline),
test=dict(
type=dataset_type,
ann_file=ann_file_test,
data_prefix=data_root_val,
filename_tmpl='{:05}.jpg',
pipeline=test_pipeline))
evaluation = dict(
interval=1, metrics=['top_k_accuracy'], start=18, gpu_collect=True)
# optimizer
optimizer = dict(
type='SGD', lr=0.06, momentum=0.9,
weight_decay=0.000001) # this lr is used for 8 gpus
optimizer_config = dict(grad_clip=dict(max_norm=40, norm_type=2))
# learning policy
lr_config = dict(
policy='step',
step=[14, 18],
warmup='linear',
warmup_by_epoch=False,
warmup_iters=16343 // 32)
total_epochs = 22
# runtime settings
checkpoint_config = dict(interval=1)
work_dir = './work_dirs/slowfast_r50_16x8x1_22e_sthv1_rgb'
load_from = 'https://download.openmmlab.com/mmaction/recognition/slowfast/slowfast_r50_8x8x1_256e_kinetics400_rgb/slowfast_r50_8x8x1_256e_kinetics400_rgb_20200716-73547d2b.pth' # noqa: E501
find_unused_parameters = False
configs/recognition/slowfast/slowfast_r50_4x16x1_256e_kinetics400_rgb.py 0 → 100644
View file @ 5b3e36dc
_base_ = [
'../../_base_/models/slowfast_r50.py', '../../_base_/default_runtime.py'
]
dataset_type = 'RawframeDataset'
data_root = 'data/kinetics400/rawframes_train'
data_root_val = 'data/kinetics400/rawframes_val'
ann_file_train = 'data/kinetics400/kinetics400_train_list_rawframes.txt'
ann_file_val = 'data/kinetics400/kinetics400_val_list_rawframes.txt'
ann_file_test = 'data/kinetics400/kinetics400_val_list_rawframes.txt'
img_norm_cfg = dict(
mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_bgr=False)
train_pipeline = [
dict(type='SampleFrames', clip_len=32, frame_interval=2, num_clips=1),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='RandomResizedCrop'),
dict(type='Resize', scale=(224, 224), keep_ratio=False),
dict(type='Flip', flip_ratio=0.5),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs', 'label'])
]
val_pipeline = [
dict(
type='SampleFrames',
clip_len=32,
frame_interval=2,
num_clips=1,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='CenterCrop', crop_size=224),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
test_pipeline = [
dict(
type='SampleFrames',
clip_len=32,
frame_interval=2,
num_clips=10,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='ThreeCrop', crop_size=256),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
data = dict(
videos_per_gpu=8,
workers_per_gpu=2,
test_dataloader=dict(videos_per_gpu=1),
train=dict(
type=dataset_type,
ann_file=ann_file_train,
data_prefix=data_root,
pipeline=train_pipeline),
val=dict(
type=dataset_type,
ann_file=ann_file_val,
data_prefix=data_root_val,
pipeline=val_pipeline),
test=dict(
type=dataset_type,
ann_file=ann_file_test,
data_prefix=data_root_val,
pipeline=test_pipeline))
evaluation = dict(
interval=5, metrics=['top_k_accuracy', 'mean_class_accuracy'])
# optimizer
optimizer = dict(
type='SGD', lr=0.1, momentum=0.9,
weight_decay=0.0001) # this lr is used for 8 gpus
optimizer_config = dict(grad_clip=dict(max_norm=40, norm_type=2))
# learning policy
lr_config = dict(
policy='CosineAnnealing',
min_lr=0,
warmup='linear',
warmup_by_epoch=True,
warmup_iters=34)
total_epochs = 256
# runtime settings
checkpoint_config = dict(interval=4)
work_dir = './work_dirs/slowfast_r50_3d_4x16x1_256e_kinetics400_rgb'
find_unused_parameters = False
configs/recognition/slowfast/slowfast_r50_8x8x1_256e_kinetics400_rgb.py 0 → 100644
View file @ 5b3e36dc
_base_ = ['./slowfast_r50_4x16x1_256e_kinetics400_rgb.py']
model = dict(
backbone=dict(
resample_rate=4, # tau
speed_ratio=4, # alpha
channel_ratio=8, # beta_inv
slow_pathway=dict(fusion_kernel=7)))
work_dir = './work_dirs/slowfast_r50_3d_8x8x1_256e_kinetics400_rgb'
configs/recognition/slowfast/slowfast_r50_8x8x1_256e_kinetics400_rgb_steplr.py 0 → 100644
View file @ 5b3e36dc
_base_ = ['./slowfast_r50_8x8x1_256e_kinetics400_rgb.py']
model = dict(backbone=dict(slow_pathway=dict(lateral_norm=True)))
lr_config = dict(
policy='step',
min_lr=0,
warmup='linear',
warmup_by_epoch=True,
warmup_iters=34,
step=[94, 154, 196])
work_dir = './work_dirs/slowfast_r50_8x8x1_256e_kinetics400_rgb_steplr'
configs/recognition/slowfast/slowfast_r50_video_4x16x1_256e_kinetics400_rgb.py 0 → 100644
View file @ 5b3e36dc
_base_ = ['./slowfast_r50_4x16x1_256e_kinetics400_rgb.py']
model = dict(
backbone=dict(
resample_rate=8, # tau
speed_ratio=8, # alpha
channel_ratio=8 # beta_inv
))
# dataset settings
dataset_type = 'VideoDataset'
data_root = 'data/kinetics400/videos_train'
data_root_val = 'data/kinetics400/videos_val'
ann_file_train = 'data/kinetics400/kinetics400_train_list_videos.txt'
ann_file_val = 'data/kinetics400/kinetics400_val_list_videos.txt'
ann_file_test = 'data/kinetics400/kinetics400_val_list_videos.txt'
img_norm_cfg = dict(
mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_bgr=False)
train_pipeline = [
dict(type='DecordInit'),
dict(type='SampleFrames', clip_len=32, frame_interval=2, num_clips=1),
dict(type='DecordDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='RandomResizedCrop'),
dict(type='Resize', scale=(224, 224), keep_ratio=False),
dict(type='Flip', flip_ratio=0.5),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs', 'label'])
]
val_pipeline = [
dict(type='DecordInit'),
dict(
type='SampleFrames',
clip_len=32,
frame_interval=2,
num_clips=1,
test_mode=True),
dict(type='DecordDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='CenterCrop', crop_size=224),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
test_pipeline = [
dict(type='DecordInit'),
dict(
type='SampleFrames',
clip_len=32,
frame_interval=2,
num_clips=10,
test_mode=True),
dict(type='DecordDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='ThreeCrop', crop_size=256),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
data = dict(
videos_per_gpu=8,
workers_per_gpu=2,
test_dataloader=dict(videos_per_gpu=1),
train=dict(
type=dataset_type,
ann_file=ann_file_train,
data_prefix=data_root,
pipeline=train_pipeline),
val=dict(
type=dataset_type,
ann_file=ann_file_val,
data_prefix=data_root_val,
pipeline=val_pipeline),
test=dict(
type=dataset_type,
ann_file=ann_file_test,
data_prefix=data_root_val,
pipeline=test_pipeline))
# runtime settings
work_dir = './work_dirs/slowfast_r50_video_3d_4x16x1_256e_kinetics400_rgb'
configs/recognition/slowfast/slowfast_r50_video_inference_4x16x1_256e_kinetics400_rgb.py 0 → 100644
View file @ 5b3e36dc
_base_ = ['../../_base_/models/slowfast_r50.py']
# dataset settings
dataset_type = 'VideoDataset'
img_norm_cfg = dict(
mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_bgr=False)
test_pipeline = [
dict(type='DecordInit'),
dict(
type='SampleFrames',
clip_len=32,
frame_interval=2,
num_clips=10,
test_mode=True),
dict(type='DecordDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='ThreeCrop', crop_size=256),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
data = dict(
videos_per_gpu=1,
workers_per_gpu=2,
test=dict(
type=dataset_type,
ann_file=None,
data_prefix=None,
pipeline=test_pipeline))
configs/recognition/slowonly/README.md 0 → 100644
View file @ 5b3e36dc
# SlowOnly
[Slowfast networks for video recognition](https://openaccess.thecvf.com/content_ICCV_2019/html/Feichtenhofer_SlowFast_Networks_for_Video_Recognition_ICCV_2019_paper.html)
<!-- [ALGORITHM] -->
## Abstract
<!-- [ABSTRACT] -->
We present SlowFast networks for video recognition. Our model involves (i) a Slow pathway, operating at low frame rate, to capture spatial semantics, and (ii) a Fast pathway, operating at high frame rate, to capture motion at fine temporal resolution. The Fast pathway can be made very lightweight by reducing its channel capacity, yet can learn useful temporal information for video recognition. Our models achieve strong performance for both action classification and detection in video, and large improvements are pin-pointed as contributions by our SlowFast concept. We report state-of-the-art accuracy on major video recognition benchmarks, Kinetics, Charades and AVA.
<!-- [IMAGE] -->
<div align=center>
<img src="https://user-images.githubusercontent.com/34324155/143044111-94676f64-7ba8-4081-9011-f8054bed7030.png" width="800"/>
</div>
## Results and Models
### Kinetics-400
| config | resolution | gpus | backbone | pretrain | top1 acc | top5 acc | inference_time(video/s) | gpu_mem(M) | ckpt | log | json |
| :-------------------------------------------------------------------------------------------------------------------------------------------------------------- | :------------: | :--: | :------: | :------: | :------: | :------: | :---------------------: | :--------: | :--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: |
| [slowonly_r50_4x16x1_256e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb.py) | short-side 256 | 8x4 | ResNet50 | None | 72.76 | 90.51 | x | 3168 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb_20200820-bea7701f.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb/20200817_001411.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb/20200817_001411.log.json) |
| [slowonly_r50_video_4x16x1_256e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_r50_video_4x16x1_256e_kinetics400_rgb.py) | short-side 320 | 8x2 | ResNet50 | None | 72.90 | 90.82 | x | 8472 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb_20201014-c9cdc656.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb_20201014.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb_20201014.json) |
| [slowonly_r50_8x8x1_256e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_rgb.py) | short-side 256 | 8x4 | ResNet50 | None | 74.42 | 91.49 | x | 5820 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_8x8x1_256e_kinetics400_rgb/slowonly_r50_256p_8x8x1_256e_kinetics400_rgb_20200820-75851a7d.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_8x8x1_256e_kinetics400_rgb/20200817_003320.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_8x8x1_256e_kinetics400_rgb/20200817_003320.log.json) |
| [slowonly_r50_4x16x1_256e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb.py) | short-side 320 | 8x2 | ResNet50 | None | 73.02 | 90.77 | 4.0 (40x3 frames) | 3168 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/slowonly_r50_4x16x1_256e_kinetics400_rgb_20200704-a69556c6.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/so_4x16.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/slowonly_r50_4x16_73.02_90.77.log.json) |
| [slowonly_r50_8x8x1_256e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_rgb.py) | short-side 320 | 8x3 | ResNet50 | None | 74.93 | 91.92 | 2.3 (80x3 frames) | 5820 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_rgb/slowonly_r50_8x8x1_256e_kinetics400_rgb_20200703-a79c555a.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_rgb/so_8x8.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_rgb/slowonly_r50_8x8_74.93_91.92.log.json) |
| [slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb.py) | short-side 320 | 8x2 | ResNet50 | ImageNet | 73.39 | 91.12 | x | 3168 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb/slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb_20200912-1e8fc736.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb/slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb_20200912.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb/slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb_20200912.json) |
| [slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb.py) | short-side 320 | 8x4 | ResNet50 | ImageNet | 75.55 | 92.04 | x | 5820 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb/slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb_20200912-3f9ce182.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb/slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb_20200912.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb/slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb_20200912.json) |
| [slowonly_nl_embedded_gaussian_r50_4x16x1_150e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_4x16x1_150e_kinetics400_rgb.py) | short-side 320 | 8x2 | ResNet50 | ImageNet | 74.54 | 91.73 | x | 4435 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_4x16x1_150e_kinetics400_rgb/slowonly_nl_embedded_gaussian_r50_4x16x1_150e_kinetics400_rgb_20210308-0d6e5a69.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_4x16x1_150e_kinetics400_rgb/20210305_152630.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_4x16x1_150e_kinetics400_rgb/20210305_152630.log.json) |
| [slowonly_nl_embedded_gaussian_r50_8x8x1_150e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_8x8x1_150e_kinetics400_rgb.py) | short-side 320 | 8x4 | ResNet50 | ImageNet | 76.07 | 92.42 | x | 8895 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_8x8x1_150e_kinetics400_rgb/slowonly_nl_embedded_gaussian_r50_8x8x1_150e_kinetics400_rgb_20210308-e8dd9e82.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_8x8x1_150e_kinetics400_rgb/20210308_212250.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_8x8x1_150e_kinetics400_rgb/20210308_212250.log.json) |
| [slowonly_r50_4x16x1_256e_kinetics400_flow](/configs/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_flow.py) | short-side 320 | 8x2 | ResNet50 | ImageNet | 61.79 | 83.62 | x | 8450 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_flow/slowonly_r50_4x16x1_256e_kinetics400_flow_20200704-decb8568.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_flow/slowonly_r50_4x16x1_256e_kinetics400_flow_61.8_83.6.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_flow/slowonly_r50_4x16x1_256e_kinetics400_flow_61.8_83.6.log.json) |
| [slowonly_r50_8x8x1_196e_kinetics400_flow](/configs/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_flow.py) | short-side 320 | 8x4 | ResNet50 | ImageNet | 65.76 | 86.25 | x | 8455 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_flow/slowonly_r50_8x8x1_256e_kinetics400_flow_20200704-6b384243.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_flow/slowonly_r50_8x8x1_196e_kinetics400_flow_65.8_86.3.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_flow/slowonly_r50_8x8x1_196e_kinetics400_flow_65.8_86.3.log.json) |
### Kinetics-400 Data Benchmark
In data benchmark, we compare two different data preprocessing methods: (1) Resize video to 340x256, (2) Resize the short edge of video to 320px, (3) Resize the short edge of video to 256px.
| config | resolution | gpus | backbone | Input | pretrain | top1 acc | top5 acc | testing protocol | ckpt | log | json |
| :--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | :------------: | :--: | :------: | :---: | :------: | :------: | :------: | :----------------: | :---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: |
| [slowonly_r50_randomresizedcrop_340x256_4x16x1_256e_kinetics400_rgb](/configs/recognition/slowonly/data_benchmark/slowonly_r50_randomresizedcrop_340x256_4x16x1_256e_kinetics400_rgb.py) | 340x256 | 8x2 | ResNet50 | 4x16 | None | 71.61 | 90.05 | 10 clips x 3 crops | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/data_benchmark/slowonly_r50_randomresizedcrop_340x256_4x16x1_256e_kinetics400_rgb/slowonly_r50_randomresizedcrop_340x256_4x16x1_256e_kinetics400_rgb_20200803-dadca1a3.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/data_benchmark/slowonly_r50_randomresizedcrop_340x256_4x16x1_256e_kinetics400_rgb/slowonly_r50_randomresizedcrop_340x256_4x16x1_256e_kinetics400_rgb_20200803.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/data_benchmark/slowonly_r50_randomresizedcrop_340x256_4x16x1_256e_kinetics400_rgb/slowonly_r50_randomresizedcrop_340x256_4x16x1_256e_kinetics400_rgb_20200803.json) |
| [slowonly_r50_randomresizedcrop_320p_4x16x1_256e_kinetics400_rgb](/configs/recognition/slowonly/data_benchmark/slowonly_r50_randomresizedcrop_320p_4x16x1_256e_kinetics400_rgb.py) | short-side 320 | 8x2 | ResNet50 | 4x16 | None | 73.02 | 90.77 | 10 clips x 3 crops | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/slowonly_r50_4x16x1_256e_kinetics400_rgb_20200704-a69556c6.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/so_4x16.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/slowonly_r50_4x16_73.02_90.77.log.json) |
| [slowonly_r50_randomresizedcrop_256p_4x16x1_256e_kinetics400_rgb](/configs/recognition/slowonly/data_benchmark/slowonly_r50_randomresizedcrop_256p_4x16x1_256e_kinetics400_rgb.py) | short-side 256 | 8x4 | ResNet50 | 4x16 | None | 72.76 | 90.51 | 10 clips x 3 crops | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb_20200820-bea7701f.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb/20200817_001411.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb/20200817_001411.log.json) |
### Kinetics-400 OmniSource Experiments
| config | resolution | backbone | pretrain | w. OmniSource | top1 acc | top5 acc | ckpt | log | json |
| :-------------------------------------------------------------------------------------------------------------------: | :------------: | :-------: | :------: | :----------------: | :------: | :------: | :--------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :----------------------------------------------------------------------------------------------------------------------: | :--------------------------------------------------------------------------------------------------------------------------------------------------: |
| [slowonly_r50_4x16x1_256e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb.py) | short-side 320 | ResNet50 | None | :x: | 73.0 | 90.8 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/slowonly_r50_4x16x1_256e_kinetics400_rgb_20200704-a69556c6.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/so_4x16.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/slowonly_r50_4x16_73.02_90.77.log.json) |
| x | x | ResNet50 | None | :heavy_check_mark: | 76.8 | 92.5 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/omni/slowonly_r50_omni_4x16x1_kinetics400_rgb_20200926-51b1f7ea.pth) | x | x |
| [slowonly_r101_8x8x1_196e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_r101_8x8x1_196e_kinetics400_rgb.py) | x | ResNet101 | None | :x: | 76.5 | 92.7 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/omni/slowonly_r101_without_omni_8x8x1_kinetics400_rgb_20200926-0c730aef.pth) | x | x |
| x | x | ResNet101 | None | :heavy_check_mark: | 80.4 | 94.4 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/omni/slowonly_r101_omni_8x8x1_kinetics400_rgb_20200926-b5dbb701.pth) | x | x |
### Kinetics-600
| config | resolution | gpus | backbone | pretrain | top1 acc | top5 acc | ckpt | log | json |
| :------------------------------------------------------------------------------------------------------------------------------ | :------------: | :--: | :------: | :------: | :------: | :------: | :------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :--------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------------: |
| [slowonly_r50_video_8x8x1_256e_kinetics600_rgb](/configs/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics600_rgb.py) | short-side 256 | 8x4 | ResNet50 | None | 77.5 | 93.7 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics600_rgb/slowonly_r50_video_8x8x1_256e_kinetics600_rgb_20201015-81e5153e.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics600_rgb/slowonly_r50_video_8x8x1_256e_kinetics600_rgb_20201015.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics600_rgb/slowonly_r50_video_8x8x1_256e_kinetics600_rgb_20201015.json) |
### Kinetics-700
| config | resolution | gpus | backbone | pretrain | top1 acc | top5 acc | ckpt | log | json |
| :------------------------------------------------------------------------------------------------------------------------------ | :------------: | :--: | :------: | :------: | :------: | :------: | :------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :--------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------------: |
| [slowonly_r50_video_8x8x1_256e_kinetics700_rgb](/configs/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics700_rgb.py) | short-side 256 | 8x4 | ResNet50 | None | 65.0 | 86.1 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics700_rgb/slowonly_r50_video_8x8x1_256e_kinetics700_rgb_20201015-9250f662.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics700_rgb/slowonly_r50_video_8x8x1_256e_kinetics700_rgb_20201015.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics700_rgb/slowonly_r50_video_8x8x1_256e_kinetics700_rgb_20201015.json) |
### GYM99
| config | resolution | gpus | backbone | pretrain | top1 acc | mean class acc | ckpt | log | json |
| :------------------------------------------------------------------------------------------------------------------------------------------------ | :------------: | :--: | :------: | :------: | :------: | :------------: | :--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: |
| [slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb](/configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb.py) | short-side 256 | 8x2 | ResNet50 | ImageNet | 79.3 | 70.2 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb_20201111-a9c34b54.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb_20201111.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb_20201111.json) |
| [slowonly_k400_pretrained_r50_4x16x1_120e_gym99_flow](/configs/recognition/slowonly/slowonly_k400_pretrained_r50_4x16x1_120e_gym99_flow.py) | short-side 256 | 8x2 | ResNet50 | Kinetics | 80.3 | 71.0 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_kinetics_pretrained_r50_4x16x1_120e_gym99_flow/slowonly_kinetics_pretrained_r50_4x16x1_120e_gym99_flow_20201111-66ecdb3c.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_kinetics_pretrained_r50_4x16x1_120e_gym99_flow/slowonly_kinetics_pretrained_r50_4x16x1_120e_gym99_flow_20201111.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_kinetics_pretrained_r50_4x16x1_120e_gym99_flow/slowonly_kinetics_pretrained_r50_4x16x1_120e_gym99_flow_20201111.json) |
| 1: 1 Fusion | | | | | 83.7 | 74.8 | | | |
### Jester
| config | resolution | gpus | backbone | pretrain | top1 acc | ckpt | log | json |
| :---------------------------------------------------------------------------------------------------------------------------------------------- | :--------: | :--: | :------: | :------: | :------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: |
| [slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb](/configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb.py) | height 100 | 8 | ResNet50 | ImageNet | 97.2 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb/slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb-b56a5389.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb/slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb/slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb.json) |
### HMDB51
| config | gpus | backbone | pretrain | top1 acc | top5 acc | gpu_mem(M) | ckpt | log | json |
| :---------------------------------------------------------------------------------------------------------------------------------------------- | :--: | :------: | :---------: | :------: | :------: | :--------: | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------: |
| [slowonly_imagenet_pretrained_r50_8x4x1_64e_hmdb51_rgb](/configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_hmdb51_rgb.py) | 8 | ResNet50 | ImageNet | 37.52 | 71.50 | 5812 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_hmdb51_rgb/slowonly_imagenet_pretrained_r50_8x4x1_64e_hmdb51_rgb_20210630-16faeb6a.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_hmdb51_rgb/20210605_185256.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_hmdb51_rgb/20210605_185256.log.json) |
| [slowonly_k400_pretrained_r50_8x4x1_40e_hmdb51_rgb](/configs/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_hmdb51_rgb.py) | 8 | ResNet50 | Kinetics400 | 65.95 | 91.05 | 5812 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_hmdb51_rgb/slowonly_k400_pretrained_r50_8x4x1_40e_hmdb51_rgb_20210630-cee5f725.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_hmdb51_rgb/20210606_010153.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_hmdb51_rgb/20210606_010153.log.json) |
### UCF101
| config | gpus | backbone | pretrain | top1 acc | top5 acc | gpu_mem(M) | ckpt | log | json |
| :---------------------------------------------------------------------------------------------------------------------------------------------- | :--: | :------: | :---------: | :------: | :------: | :--------: | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------: |
| [slowonly_imagenet_pretrained_r50_8x4x1_64e_ucf101_rgb](/configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_ucf101_rgb.py) | 8 | ResNet50 | ImageNet | 71.35 | 89.35 | 5812 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_ucf101_rgb/slowonly_imagenet_pretrained_r50_8x4x1_64e_ucf101_rgb_20210630-181e1661.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_ucf101_rgb/20210605_213503.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_ucf101_rgb/20210605_213503.log.json) |
| [slowonly_k400_pretrained_r50_8x4x1_40e_ucf101_rgb](/configs/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_ucf101_rgb.py) | 8 | ResNet50 | Kinetics400 | 92.78 | 99.42 | 5812 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_ucf101_rgb/slowonly_k400_pretrained_r50_8x4x1_40e_ucf101_rgb_20210630-ee8c850f.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_ucf101_rgb/20210606_010231.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_ucf101_rgb/20210606_010231.log.json) |
### Something-Something V1
| config | gpus | backbone | pretrain | top1 acc | top5 acc | gpu_mem(M) | ckpt | log | json |
| :-------------------------------------------------------------------------------------------------------------------------------------------- | :--: | :------: | :------: | :------: | :------: | :--------: | :--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: |
| [slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb](/configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb.py) | 8 | ResNet50 | ImageNet | 47.76 | 77.49 | 7759 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb/slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb_20211202-d034ff12.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb/slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb/slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb.json) |
:::{note}
1. The **gpus** indicates the number of gpu we used to get the checkpoint. It is noteworthy that the configs we provide are used for 8 gpus as default.
According to the [Linear Scaling Rule](https://arxiv.org/abs/1706.02677), you may set the learning rate proportional to the batch size if you use different GPUs or videos per GPU,
e.g., lr=0.01 for 4 GPUs x 2 video/gpu and lr=0.08 for 16 GPUs x 4 video/gpu.
2. The **inference_time** is got by this [benchmark script](/tools/analysis/benchmark.py), where we use the sampling frames strategy of the test setting and only care about the model inference time, not including the IO time and pre-processing time. For each setting, we use 1 gpu and set batch size (videos per gpu) to 1 to calculate the inference time.
3. The validation set of Kinetics400 we used consists of 19796 videos. These videos are available at [Kinetics400-Validation](https://mycuhk-my.sharepoint.com/:u:/g/personal/1155136485_link_cuhk_edu_hk/EbXw2WX94J1Hunyt3MWNDJUBz-nHvQYhO9pvKqm6g39PMA?e=a9QldB). The corresponding [data list](https://download.openmmlab.com/mmaction/dataset/k400_val/kinetics_val_list.txt) (each line is of the format 'video_id, num_frames, label_index') and the [label map](https://download.openmmlab.com/mmaction/dataset/k400_val/kinetics_class2ind.txt) are also available.
:::
For more details on data preparation, you can refer to corresponding parts in [Data Preparation](/docs/en/data_preparation.md).
## Train
You can use the following command to train a model.
```shell
python tools/train.py ${CONFIG_FILE} [optional arguments]
```
Example: train SlowOnly model on Kinetics-400 dataset in a deterministic option with periodic validation.
```shell
python tools/train.py configs/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb.py \
--work-dir work_dirs/slowonly_r50_4x16x1_256e_kinetics400_rgb \
--validate --seed 0 --deterministic
```
For more details, you can refer to **Training setting** part in [getting_started](/docs/en/getting_started.md#training-setting).
## Test
You can use the following command to test a model.
```shell
python tools/test.py ${CONFIG_FILE} ${CHECKPOINT_FILE} [optional arguments]
```
Example: test SlowOnly model on Kinetics-400 dataset and dump the result to a json file.
```shell
python tools/test.py configs/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb.py \
checkpoints/SOME_CHECKPOINT.pth --eval top_k_accuracy mean_class_accuracy \
--out result.json --average-clips=prob
```
For more details, you can refer to **Test a dataset** part in [getting_started](/docs/en/getting_started.md#test-a-dataset).
## Citation
```BibTeX
@inproceedings{feichtenhofer2019slowfast,
title={Slowfast networks for video recognition},
author={Feichtenhofer, Christoph and Fan, Haoqi and Malik, Jitendra and He, Kaiming},
booktitle={Proceedings of the IEEE international conference on computer vision},
pages={6202--6211},
year={2019}
}
```
configs/recognition/slowonly/README_zh-CN.md 0 → 100644
View file @ 5b3e36dc
# SlowOnly
## 简介
<!-- [ALGORITHM] -->
```BibTeX
@inproceedings{feichtenhofer2019slowfast,
title={Slowfast networks for video recognition},
author={Feichtenhofer, Christoph and Fan, Haoqi and Malik, Jitendra and He, Kaiming},
booktitle={Proceedings of the IEEE international conference on computer vision},
pages={6202--6211},
year={2019}
}
```
## 模型库
### Kinetics-400
| 配置文件 | 分辨率 | GPU 数量 | 主干网络 | 预训练 | top1 准确率 | top5 准确率 | 推理时间 (video/s) | GPU 显存占用 (M) | ckpt | log | json |
| :-------------------------------------------------------------------------------------------------------------------------------------------------------------- | :------: | :------: | :------: | :------: | :---------: | :---------: | :----------------: | :--------------: | :--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: |
| [slowonly_r50_4x16x1_256e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb.py) | 短边 256 | 8x4 | ResNet50 | None | 72.76 | 90.51 | x | 3168 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb_20200820-bea7701f.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb/20200817_001411.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb/20200817_001411.log.json) |
| [slowonly_r50_video_4x16x1_256e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_r50_video_4x16x1_256e_kinetics400_rgb.py) | 短边 320 | 8x2 | ResNet50 | None | 72.90 | 90.82 | x | 8472 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb_20201014-c9cdc656.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb_20201014.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb_20201014.json) |
| [slowonly_r50_8x8x1_256e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_rgb.py) | 短边 256 | 8x4 | ResNet50 | None | 74.42 | 91.49 | x | 5820 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_8x8x1_256e_kinetics400_rgb/slowonly_r50_256p_8x8x1_256e_kinetics400_rgb_20200820-75851a7d.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_8x8x1_256e_kinetics400_rgb/20200817_003320.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_8x8x1_256e_kinetics400_rgb/20200817_003320.log.json) |
| [slowonly_r50_4x16x1_256e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb.py) | 短边 320 | 8x2 | ResNet50 | None | 73.02 | 90.77 | 4.0 (40x3 frames) | 3168 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/slowonly_r50_4x16x1_256e_kinetics400_rgb_20200704-a69556c6.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/so_4x16.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/slowonly_r50_4x16_73.02_90.77.log.json) |
| [slowonly_r50_8x8x1_256e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_rgb.py) | 短边 320 | 8x3 | ResNet50 | None | 74.93 | 91.92 | 2.3 (80x3 frames) | 5820 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_rgb/slowonly_r50_8x8x1_256e_kinetics400_rgb_20200703-a79c555a.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_rgb/so_8x8.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_rgb/slowonly_r50_8x8_74.93_91.92.log.json) |
| [slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb.py) | 短边 320 | 8x2 | ResNet50 | ImageNet | 73.39 | 91.12 | x | 3168 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb/slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb_20200912-1e8fc736.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb/slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb_20200912.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb/slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb_20200912.json) |
| [slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb.py) | 短边 320 | 8x4 | ResNet50 | ImageNet | 75.55 | 92.04 | x | 5820 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb/slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb_20200912-3f9ce182.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb/slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb_20200912.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb/slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb_20200912.json) |
| [slowonly_nl_embedded_gaussian_r50_4x16x1_150e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_4x16x1_150e_kinetics400_rgb.py) | 短边 320 | 8x2 | ResNet50 | ImageNet | 74.54 | 91.73 | x | 4435 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_4x16x1_150e_kinetics400_rgb/slowonly_nl_embedded_gaussian_r50_4x16x1_150e_kinetics400_rgb_20210308-0d6e5a69.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_4x16x1_150e_kinetics400_rgb/20210305_152630.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_4x16x1_150e_kinetics400_rgb/20210305_152630.log.json) |
| [slowonly_nl_embedded_gaussian_r50_8x8x1_150e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_8x8x1_150e_kinetics400_rgb.py) | 短边 320 | 8x4 | ResNet50 | ImageNet | 76.07 | 92.42 | x | 8895 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_8x8x1_150e_kinetics400_rgb/slowonly_nl_embedded_gaussian_r50_8x8x1_150e_kinetics400_rgb_20210308-e8dd9e82.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_8x8x1_150e_kinetics400_rgb/20210308_212250.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_8x8x1_150e_kinetics400_rgb/20210308_212250.log.json) |
| [slowonly_r50_4x16x1_256e_kinetics400_flow](/configs/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_flow.py) | 短边 320 | 8x2 | ResNet50 | ImageNet | 61.79 | 83.62 | x | 8450 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_flow/slowonly_r50_4x16x1_256e_kinetics400_flow_20200704-decb8568.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_flow/slowonly_r50_4x16x1_256e_kinetics400_flow_61.8_83.6.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_flow/slowonly_r50_4x16x1_256e_kinetics400_flow_61.8_83.6.log.json) |
| [slowonly_r50_8x8x1_196e_kinetics400_flow](/configs/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_flow.py) | 短边 320 | 8x4 | ResNet50 | ImageNet | 65.76 | 86.25 | x | 8455 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_flow/slowonly_r50_8x8x1_256e_kinetics400_flow_20200704-6b384243.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_flow/slowonly_r50_8x8x1_196e_kinetics400_flow_65.8_86.3.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_flow/slowonly_r50_8x8x1_196e_kinetics400_flow_65.8_86.3.log.json) |
### Kinetics-400 数据基准测试
在数据基准测试中，比较两种不同的数据预处理方法 (1) 视频分辨率为 340x256, (2) 视频分辨率为短边 320px, (3) 视频分辨率为短边 256px.
| 配置文件 | 分辨率 | GPU 数量 | 主干网络 | 输入 | 预训练 | top1 准确率 | top5 准确率 | 测试方案 | ckpt | log | json |
| :--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | :------: | :------: | :------: | :--: | :----: | :---------: | :---------: | :----------------: | :---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: |
| [slowonly_r50_randomresizedcrop_340x256_4x16x1_256e_kinetics400_rgb](/configs/recognition/slowonly/data_benchmark/slowonly_r50_randomresizedcrop_340x256_4x16x1_256e_kinetics400_rgb.py) | 340x256 | 8x2 | ResNet50 | 4x16 | None | 71.61 | 90.05 | 10 clips x 3 crops | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/data_benchmark/slowonly_r50_randomresizedcrop_340x256_4x16x1_256e_kinetics400_rgb/slowonly_r50_randomresizedcrop_340x256_4x16x1_256e_kinetics400_rgb_20200803-dadca1a3.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/data_benchmark/slowonly_r50_randomresizedcrop_340x256_4x16x1_256e_kinetics400_rgb/slowonly_r50_randomresizedcrop_340x256_4x16x1_256e_kinetics400_rgb_20200803.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/data_benchmark/slowonly_r50_randomresizedcrop_340x256_4x16x1_256e_kinetics400_rgb/slowonly_r50_randomresizedcrop_340x256_4x16x1_256e_kinetics400_rgb_20200803.json) |
| [slowonly_r50_randomresizedcrop_320p_4x16x1_256e_kinetics400_rgb](/configs/recognition/slowonly/data_benchmark/slowonly_r50_randomresizedcrop_320p_4x16x1_256e_kinetics400_rgb.py) | 短边 320 | 8x2 | ResNet50 | 4x16 | None | 73.02 | 90.77 | 10 clips x 3 crops | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/slowonly_r50_4x16x1_256e_kinetics400_rgb_20200704-a69556c6.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/so_4x16.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/slowonly_r50_4x16_73.02_90.77.log.json) |
| [slowonly_r50_randomresizedcrop_256p_4x16x1_256e_kinetics400_rgb](/configs/recognition/slowonly/data_benchmark/slowonly_r50_randomresizedcrop_256p_4x16x1_256e_kinetics400_rgb.py) | 短边 256 | 8x4 | ResNet50 | 4x16 | None | 72.76 | 90.51 | 10 clips x 3 crops | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb_20200820-bea7701f.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb/20200817_001411.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb/20200817_001411.log.json) |
### Kinetics-400 OmniSource Experiments
| 配置文件 | 分辨率 | 主干网络 | 预训练 | w. OmniSource | top1 准确率 | top5 准确率 | ckpt | log | json |
| :-------------------------------------------------------------------------------------------------------------------: | :------: | :-------: | :----: | :----------------: | :---------: | :---------: | :--------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :----------------------------------------------------------------------------------------------------------------------: | :--------------------------------------------------------------------------------------------------------------------------------------------------: |
| [slowonly_r50_4x16x1_256e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb.py) | 短边 320 | ResNet50 | None | :x: | 73.0 | 90.8 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/slowonly_r50_4x16x1_256e_kinetics400_rgb_20200704-a69556c6.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/so_4x16.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/slowonly_r50_4x16_73.02_90.77.log.json) |
| x | x | ResNet50 | None | :heavy_check_mark: | 76.8 | 92.5 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/omni/slowonly_r50_omni_4x16x1_kinetics400_rgb_20200926-51b1f7ea.pth) | x | x |
| [slowonly_r101_8x8x1_196e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_r101_8x8x1_196e_kinetics400_rgb.py) | x | ResNet101 | None | :x: | 76.5 | 92.7 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/omni/slowonly_r101_without_omni_8x8x1_kinetics400_rgb_20200926-0c730aef.pth) | x | x |
| x | x | ResNet101 | None | :heavy_check_mark: | 80.4 | 94.4 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/omni/slowonly_r101_omni_8x8x1_kinetics400_rgb_20200926-b5dbb701.pth) | x | x |
### Kinetics-600
| 配置文件 | 分辨率 | GPU 数量 | 主干网络 | 预训练 | top1 准确率 | top5 准确率 | ckpt | log | json |
| :------------------------------------------------------------------------------------------------------------------------------ | :------: | :------: | :------: | :----: | :---------: | :---------: | :------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :--------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------------: |
| [slowonly_r50_video_8x8x1_256e_kinetics600_rgb](/configs/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics600_rgb.py) | 短边 256 | 8x4 | ResNet50 | None | 77.5 | 93.7 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics600_rgb/slowonly_r50_video_8x8x1_256e_kinetics600_rgb_20201015-81e5153e.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics600_rgb/slowonly_r50_video_8x8x1_256e_kinetics600_rgb_20201015.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics600_rgb/slowonly_r50_video_8x8x1_256e_kinetics600_rgb_20201015.json) |
### Kinetics-700
| 配置文件 | 分辨率 | GPU 数量 | 主干网络 | 预训练 | top1 准确率 | top5 准确率 | ckpt | log | json |
| :------------------------------------------------------------------------------------------------------------------------------ | :------: | :------: | :------: | :----: | :---------: | :---------: | :------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :--------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------------: |
| [slowonly_r50_video_8x8x1_256e_kinetics700_rgb](/configs/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics700_rgb.py) | 短边 256 | 8x4 | ResNet50 | None | 65.0 | 86.1 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics700_rgb/slowonly_r50_video_8x8x1_256e_kinetics700_rgb_20201015-9250f662.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics700_rgb/slowonly_r50_video_8x8x1_256e_kinetics700_rgb_20201015.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics700_rgb/slowonly_r50_video_8x8x1_256e_kinetics700_rgb_20201015.json) |
### GYM99
| 配置文件 | 分辨率 | GPU 数量 | 主干网络 | 预训练 | top1 准确率 | 类别平均准确率 | ckpt | log | json |
| :------------------------------------------------------------------------------------------------------------------------------------------------ | :------: | :------: | :------: | :------: | :---------: | :------------: | :--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: |
| [slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb](/configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb.py) | 短边 256 | 8x2 | ResNet50 | ImageNet | 79.3 | 70.2 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb_20201111-a9c34b54.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb_20201111.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb_20201111.json) |
| [slowonly_kinetics_pretrained_r50_4x16x1_120e_gym99_flow](/configs/recognition/slowonly/slowonly_k400_pretrained_r50_4x16x1_120e_gym99_flow.py) | 短边 256 | 8x2 | ResNet50 | Kinetics | 80.3 | 71.0 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_kinetics_pretrained_r50_4x16x1_120e_gym99_flow/slowonly_kinetics_pretrained_r50_4x16x1_120e_gym99_flow_20201111-66ecdb3c.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_kinetics_pretrained_r50_4x16x1_120e_gym99_flow/slowonly_kinetics_pretrained_r50_4x16x1_120e_gym99_flow_20201111.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_kinetics_pretrained_r50_4x16x1_120e_gym99_flow/slowonly_kinetics_pretrained_r50_4x16x1_120e_gym99_flow_20201111.json) |
| 1: 1 融合 | | | | | 83.7 | 74.8 | | | |
### Jester
| 配置文件 | 分辨率 | GPU 数量 | 主干网络 | 预训练 | top1 准确率 | ckpt | log | json |
| :---------------------------------------------------------------------------------------------------------------------------------------------- | :----: | :------: | :------: | :------: | :---------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: |
| [slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb](/configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb.py) | 高 100 | 8 | ResNet50 | ImageNet | 97.2 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb/slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb-b56a5389.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb/slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb/slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb.json) |
### HMDB51
| 配置文件 | GPU 数量 | 主干网络 | 预训练 | top1 准确率 | top5 准确率 | GPU 显存占用 (M) | ckpt | log | json |
| :---------------------------------------------------------------------------------------------------------------------------------------------- | :------: | :------: | :---------: | :---------: | :---------: | :--------------: | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------: |
| [slowonly_imagenet_pretrained_r50_8x4x1_64e_hmdb51_rgb](/configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_hmdb51_rgb.py) | 8 | ResNet50 | ImageNet | 37.52 | 71.50 | 5812 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_hmdb51_rgb/slowonly_imagenet_pretrained_r50_8x4x1_64e_hmdb51_rgb_20210630-16faeb6a.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_hmdb51_rgb/20210605_185256.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_hmdb51_rgb/20210605_185256.log.json) |
| [slowonly_k400_pretrained_r50_8x4x1_40e_hmdb51_rgb](/configs/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_hmdb51_rgb.py) | 8 | ResNet50 | Kinetics400 | 65.95 | 91.05 | 5812 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_hmdb51_rgb/slowonly_k400_pretrained_r50_8x4x1_40e_hmdb51_rgb_20210630-cee5f725.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_hmdb51_rgb/20210606_010153.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_hmdb51_rgb/20210606_010153.log.json) |
### UCF101
| 配置文件 | GPU 数量 | 主干网络 | 预训练 | top1 准确率 | top5 准确率 | GPU 显存占用 (M) | ckpt | log | json |
| :---------------------------------------------------------------------------------------------------------------------------------------------- | :------: | :------: | :---------: | :---------: | :---------: | :--------------: | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------: |
| [slowonly_imagenet_pretrained_r50_8x4x1_64e_ucf101_rgb](/configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_ucf101_rgb.py) | 8 | ResNet50 | ImageNet | 71.35 | 89.35 | 5812 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_ucf101_rgb/slowonly_imagenet_pretrained_r50_8x4x1_64e_ucf101_rgb_20210630-181e1661.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_ucf101_rgb/20210605_213503.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_ucf101_rgb/20210605_213503.log.json) |
| [slowonly_k400_pretrained_r50_8x4x1_40e_ucf101_rgb](/configs/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_ucf101_rgb.py) | 8 | ResNet50 | Kinetics400 | 92.78 | 99.42 | 5812 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_ucf101_rgb/slowonly_k400_pretrained_r50_8x4x1_40e_ucf101_rgb_20210630-ee8c850f.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_ucf101_rgb/20210606_010231.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_ucf101_rgb/20210606_010231.log.json) |
### Something-Something V1
| 配置文件 | GPU 数量 | 主干网络 | 预训练 | top1 准确率 | top5 准确率 | GPU 显存占用 (M) | ckpt | log | json |
| :-------------------------------------------------------------------------------------------------------------------------------------------- | :------: | :------: | :------: | :---------: | :---------: | :--------------: | :--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: |
| [slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb](/configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb.py) | 8 | ResNet50 | ImageNet | 47.76 | 77.49 | 7759 | [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb/slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb_20211202-d034ff12.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb/slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb.log) | [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb/slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb.json) |
注：
1. 这里的 **GPU 数量** 指的是得到模型权重文件对应的 GPU 个数。默认地，MMAction2 所提供的配置文件对应使用 8 块 GPU 进行训练的情况。
依据 [线性缩放规则](https://arxiv.org/abs/1706.02677)，当用户使用不同数量的 GPU 或者每块 GPU 处理不同视频个数时，需要根据批大小等比例地调节学习率。
如，lr=0.01 对应 4 GPUs x 2 video/gpu，以及 lr=0.08 对应 16 GPUs x 4 video/gpu。
2. 这里的 **推理时间** 是根据 [基准测试脚本](/tools/analysis/benchmark.py) 获得的，采用测试时的采帧策略，且只考虑模型的推理时间，
并不包括 IO 时间以及预处理时间。对于每个配置，MMAction2 使用 1 块 GPU 并设置批大小（每块 GPU 处理的视频个数）为 1 来计算推理时间。
3. 我们使用的 Kinetics400 验证集包含 19796 个视频，用户可以从 [验证集视频](https://mycuhk-my.sharepoint.com/:u:/g/personal/1155136485_link_cuhk_edu_hk/EbXw2WX94J1Hunyt3MWNDJUBz-nHvQYhO9pvKqm6g39PMA?e=a9QldB) 下载这些视频。同时也提供了对应的 [数据列表](https://download.openmmlab.com/mmaction/dataset/k400_val/kinetics_val_list.txt) （每行格式为：视频 ID，视频帧数目，类别序号）以及 [标签映射](https://download.openmmlab.com/mmaction/dataset/k400_val/kinetics_class2ind.txt) （类别序号到类别名称）。
对于数据集准备的细节，用户可参考 [数据集准备文档](/docs/zh_cn/data_preparation.md) 中的 Kinetics400 部分。
## 如何训练
用户可以使用以下指令进行模型训练。
```shell
python tools/train.py ${CONFIG_FILE} [optional arguments]
```
例如：以一个确定性的训练方式，辅以定期的验证过程进行 SlowOnly 模型在 Kinetics400 数据集上的训练。
```shell
python tools/train.py configs/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb.py \
--work-dir work_dirs/slowonly_r50_4x16x1_256e_kinetics400_rgb \
--validate --seed 0 --deterministic
```
更多训练细节，可参考 [基础教程](/docs/zh_cn/getting_started.md#训练配置) 中的 **训练配置** 部分。
## 如何测试
用户可以使用以下指令进行模型测试。
```shell
python tools/test.py ${CONFIG_FILE} ${CHECKPOINT_FILE} [optional arguments]
```
例如：在 Kinetics400 数据集上测试 SlowOnly 模型，并将结果导出为一个 json 文件。
```shell
python tools/test.py configs/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb.py \
checkpoints/SOME_CHECKPOINT.pth --eval top_k_accuracy mean_class_accuracy \
--out result.json --average-clips=prob
```
更多测试细节，可参考 [基础教程](/docs/zh_cn/getting_started.md#测试某个数据集) 中的 **测试某个数据集** 部分。
configs/recognition/slowonly/data_benchmark/slowonly_r50_randomresizedcrop_256p_4x16x1_256e_kinetics400_rgb.py 0 → 100644
View file @ 5b3e36dc
model = dict(
type='Recognizer3D',
backbone=dict(
type='ResNet3dSlowOnly',
depth=50,
pretrained=None,
lateral=False,
conv1_kernel=(1, 7, 7),
conv1_stride_t=1,
pool1_stride_t=1,
inflate=(0, 0, 1, 1),
norm_eval=False),
cls_head=dict(
type='I3DHead',
in_channels=2048,
num_classes=400,
spatial_type='avg',
dropout_ratio=0.5),
train_cfg=None,
test_cfg=dict(average_clips='prob'))
dataset_type = 'RawframeDataset'
data_root = 'data/kinetics400/rawframes_train_256p'
data_root_val = 'data/kinetics400/rawframes_val_256p'
ann_file_train = 'data/kinetics400/kinetics400_train_list_rawframes_256p.txt'
ann_file_val = 'data/kinetics400/kinetics400_val_list_rawframes_256p.txt'
ann_file_test = 'data/kinetics400/kinetics400_val_list_rawframes_256p.txt'
img_norm_cfg = dict(
mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_bgr=False)
train_pipeline = [
dict(type='SampleFrames', clip_len=4, frame_interval=16, num_clips=1),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='RandomResizedCrop'),
dict(type='Resize', scale=(224, 224), keep_ratio=False),
dict(type='Flip', flip_ratio=0.5),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs', 'label'])
]
val_pipeline = [
dict(
type='SampleFrames',
clip_len=4,
frame_interval=16,
num_clips=1,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='CenterCrop', crop_size=224),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
test_pipeline = [
dict(
type='SampleFrames',
clip_len=4,
frame_interval=16,
num_clips=10,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='ThreeCrop', crop_size=256),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
data = dict(
videos_per_gpu=16,
workers_per_gpu=2,
test_dataloader=dict(videos_per_gpu=1),
train=dict(
type=dataset_type,
ann_file=ann_file_train,
data_prefix=data_root,
pipeline=train_pipeline),
val=dict(
type=dataset_type,
ann_file=ann_file_val,
data_prefix=data_root_val,
pipeline=val_pipeline),
test=dict(
type=dataset_type,
ann_file=ann_file_test,
data_prefix=data_root_val,
pipeline=test_pipeline))
# optimizer
optimizer = dict(
type='SGD', lr=0.6, momentum=0.9,
weight_decay=0.0001) # this lr is used for 8 gpus
optimizer_config = dict(grad_clip=dict(max_norm=40, norm_type=2))
# learning policy
lr_config = dict(policy='CosineAnnealing', min_lr=0)
total_epochs = 256
checkpoint_config = dict(interval=4)
workflow = [('train', 1)]
evaluation = dict(
interval=5, metrics=['top_k_accuracy', 'mean_class_accuracy'])
log_config = dict(
interval=20,
hooks=[
dict(type='TextLoggerHook'),
# dict(type='TensorboardLoggerHook'),
])
dist_params = dict(backend='nccl')
log_level = 'INFO'
work_dir = ('./work_dirs/slowonly_r50_randomresizedcrop_256p_4x16x1'
'_256e_kinetics400_rgb')
load_from = None
resume_from = None
find_unused_parameters = False
configs/recognition/slowonly/data_benchmark/slowonly_r50_randomresizedcrop_320p_4x16x1_256e_kinetics400_rgb.py 0 → 100644
View file @ 5b3e36dc
model = dict(
type='Recognizer3D',
backbone=dict(
type='ResNet3dSlowOnly',
depth=50,
pretrained=None,
lateral=False,
conv1_kernel=(1, 7, 7),
conv1_stride_t=1,
pool1_stride_t=1,
inflate=(0, 0, 1, 1),
norm_eval=False),
cls_head=dict(
type='I3DHead',
in_channels=2048,
num_classes=400,
spatial_type='avg',
dropout_ratio=0.5),
train_cfg=None,
test_cfg=dict(average_clips='prob'))
dataset_type = 'RawframeDataset'
data_root = 'data/kinetics400/rawframes_train_320p'
data_root_val = 'data/kinetics400/rawframes_val_320p'
ann_file_train = 'data/kinetics400/kinetics400_train_list_rawframes_320p.txt'
ann_file_val = 'data/kinetics400/kinetics400_val_list_rawframes_320p.txt'
ann_file_test = 'data/kinetics400/kinetics400_val_list_rawframes_320p.txt'
img_norm_cfg = dict(
mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_bgr=False)
train_pipeline = [
dict(type='SampleFrames', clip_len=4, frame_interval=16, num_clips=1),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='RandomResizedCrop'),
dict(type='Resize', scale=(224, 224), keep_ratio=False),
dict(type='Flip', flip_ratio=0.5),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs', 'label'])
]
val_pipeline = [
dict(
type='SampleFrames',
clip_len=4,
frame_interval=16,
num_clips=1,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='CenterCrop', crop_size=224),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
test_pipeline = [
dict(
type='SampleFrames',
clip_len=4,
frame_interval=16,
num_clips=10,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='ThreeCrop', crop_size=256),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
data = dict(
videos_per_gpu=16,
workers_per_gpu=2,
test_dataloader=dict(videos_per_gpu=1),
train=dict(
type=dataset_type,
ann_file=ann_file_train,
data_prefix=data_root,
pipeline=train_pipeline),
val=dict(
type=dataset_type,
ann_file=ann_file_val,
data_prefix=data_root_val,
pipeline=val_pipeline),
test=dict(
type=dataset_type,
ann_file=ann_file_test,
data_prefix=data_root_val,
pipeline=test_pipeline))
# optimizer
optimizer = dict(
type='SGD', lr=0.6, momentum=0.9,
weight_decay=0.0001) # this lr is used for 8 gpus
optimizer_config = dict(grad_clip=dict(max_norm=40, norm_type=2))
# learning policy
lr_config = dict(policy='CosineAnnealing', min_lr=0)
total_epochs = 256
checkpoint_config = dict(interval=4)
workflow = [('train', 1)]
evaluation = dict(
interval=5, metrics=['top_k_accuracy', 'mean_class_accuracy'])
log_config = dict(
interval=20,
hooks=[
dict(type='TextLoggerHook'),
# dict(type='TensorboardLoggerHook'),
])
dist_params = dict(backend='nccl')
log_level = 'INFO'
work_dir = ('./work_dirs/slowonly_r50_randomresizedcrop_320p_4x16x1'
'_256e_kinetics400_rgb')
load_from = None
resume_from = None
find_unused_parameters = False
configs/recognition/slowonly/data_benchmark/slowonly_r50_randomresizedcrop_340x256_4x16x1_256e_kinetics400_rgb.py 0 → 100644
View file @ 5b3e36dc
model = dict(
type='Recognizer3D',
backbone=dict(
type='ResNet3dSlowOnly',
depth=50,
pretrained=None,
lateral=False,
conv1_kernel=(1, 7, 7),
conv1_stride_t=1,
pool1_stride_t=1,
inflate=(0, 0, 1, 1),
norm_eval=False),
cls_head=dict(
type='I3DHead',
in_channels=2048,
num_classes=400,
spatial_type='avg',
dropout_ratio=0.5),
train_cfg=None,
test_cfg=dict(average_clips='prob'))
dataset_type = 'RawframeDataset'
data_root = 'data/kinetics400/rawframes_train'
data_root_val = 'data/kinetics400/rawframes_val'
ann_file_train = 'data/kinetics400/kinetics400_train_list_rawframes.txt'
ann_file_val = 'data/kinetics400/kinetics400_val_list_rawframes.txt'
ann_file_test = 'data/kinetics400/kinetics400_val_list_rawframes.txt'
img_norm_cfg = dict(
mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_bgr=False)
train_pipeline = [
dict(type='SampleFrames', clip_len=4, frame_interval=16, num_clips=1),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='RandomResizedCrop'),
dict(type='Resize', scale=(224, 224), keep_ratio=False),
dict(type='Flip', flip_ratio=0.5),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs', 'label'])
]
val_pipeline = [
dict(
type='SampleFrames',
clip_len=4,
frame_interval=16,
num_clips=1,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='CenterCrop', crop_size=224),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
test_pipeline = [
dict(
type='SampleFrames',
clip_len=4,
frame_interval=16,
num_clips=10,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='ThreeCrop', crop_size=256),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
data = dict(
videos_per_gpu=16,
workers_per_gpu=2,
test_dataloader=dict(videos_per_gpu=1),
train=dict(
type=dataset_type,
ann_file=ann_file_train,
data_prefix=data_root,
pipeline=train_pipeline),
val=dict(
type=dataset_type,
ann_file=ann_file_val,
data_prefix=data_root_val,
pipeline=val_pipeline),
test=dict(
type=dataset_type,
ann_file=ann_file_test,
data_prefix=data_root_val,
pipeline=test_pipeline))
# optimizer
optimizer = dict(
type='SGD', lr=0.6, momentum=0.9,
weight_decay=0.0001) # this lr is used for 8 gpus
optimizer_config = dict(grad_clip=dict(max_norm=40, norm_type=2))
# learning policy
lr_config = dict(policy='CosineAnnealing', min_lr=0)
total_epochs = 256
checkpoint_config = dict(interval=4)
workflow = [('train', 1)]
evaluation = dict(
interval=5, metrics=['top_k_accuracy', 'mean_class_accuracy'])
log_config = dict(
interval=20,
hooks=[
dict(type='TextLoggerHook'),
# dict(type='TensorboardLoggerHook'),
])
dist_params = dict(backend='nccl')
log_level = 'INFO'
work_dir = ('slowonly_r50_randomresizedcrop_320p_4x16x1'
'_256e_kinetics400_rgb')
load_from = None
resume_from = None
find_unused_parameters = False
configs/recognition/slowonly/metafile.yml 0 → 100644
View file @ 5b3e36dc
Collections:
- Name: SlowOnly
README: configs/recognition/slowonly/README.md
Paper:
URL: https://arxiv.org/abs/1812.03982
Title: SlowFast Networks for Video Recognition
Models:
- Config: configs/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet50
Batch Size: 8
Epochs: 256
FLOPs: 27430649856
Parameters: 32454096
Pretrained: None
Resolution: short-side 320
Training Data: Kinetics-400
Modality: RGB
Name: slowonly_r50_omnisource_4x16x1_256e_kinetics400_rgb
Converted From:
Weights: https://open-mmlab.s3.ap-northeast-2.amazonaws.com/mmaction/models/kinetics400/omnisource/slowonly_OmniSource_kinetics400_se_rgb_r50_seg1_4x16_scratch-71f7b8ee.pth
Code: https://github.com/open-mmlab/mmaction/
Results:
- Dataset: Kinetics-400
Metrics:
Top 1 Accuracy: 76.8
Top 5 Accuracy: 92.5
Task: Action Recognition
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/omni/slowonly_r50_omni_4x16x1_kinetics400_rgb_20200926-51b1f7ea.pth
- Config: configs/recognition/slowonly/slowonly_r101_8x8x1_196e_kinetics400_rgb.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet101
Batch Size: 8
Epochs: 196
FLOPs: 112063447040
Parameters: 60359120
Pretrained: None
Resolution: short-side 320
Training Data: Kinetics-400
Modality: RGB
Name: slowonly_r101_8x8x1_196e_kinetics400_rgb
Results:
- Dataset: Kinetics-400
Metrics:
Top 1 Accuracy: 76.5
Top 5 Accuracy: 92.7
Task: Action Recognition
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/omni/slowonly_r101_without_omni_8x8x1_kinetics400_rgb_20200926-0c730aef.pth
- Config: configs/recognition/slowonly/slowonly_r101_8x8x1_196e_kinetics400_rgb.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet101
Batch Size: 8
Epochs: 196
FLOPs: 112063447040
Parameters: 60359120
Pretrained: None
Resolution: short-side 320
Training Data: Kinetics-400
Modality: RGB
Name: slowonly_r101_omnisource_8x8x1_196e_kinetics400_rgb
Converted From:
Weights: https://open-mmlab.s3.ap-northeast-2.amazonaws.com/mmaction/models/kinetics400/omnisource/slowonly_OmniSource_kinetics400_se_rgb_r101_seg1_8x8_scratch-2f838cb0.pth
Code: https://github.com/open-mmlab/mmaction/
Results:
- Dataset: Kinetics-400
Metrics:
Top 1 Accuracy: 80.4
Top 5 Accuracy: 94.4
Task: Action Recognition
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/omni/slowonly_r101_omni_8x8x1_kinetics400_rgb_20200926-b5dbb701.pth
- Config: configs/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet50
Batch Size: 8
Epochs: 256
FLOPs: 27430649856
Parameters: 32454096
Pretrained: None
Resolution: short-side 256
Training Data: Kinetics-400
Training Resources: 32 GPUs
Modality: RGB
Name: slowonly_r50_4x16x1_256e_kinetics400_rgb
Results:
- Dataset: Kinetics-400
Metrics:
Top 1 Accuracy: 72.76
Top 5 Accuracy: 90.51
Task: Action Recognition
Training Json Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb/20200817_001411.log.json
Training Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb/20200817_001411.log
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb_20200820-bea7701f.pth
- Config: configs/recognition/slowonly/slowonly_r50_video_4x16x1_256e_kinetics400_rgb.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet50
Batch Size: 24
Epochs: 256
FLOPs: 27430649856
Parameters: 32454096
Pretrained: None
Resolution: short-side 320
Training Data: Kinetics-400
Training Resources: 16 GPUs
Modality: RGB
Name: slowonly_r50_video_4x16x1_256e_kinetics400_rgb
Results:
- Dataset: Kinetics-400
Metrics:
Top 1 Accuracy: 72.9
Top 5 Accuracy: 90.82
Task: Action Recognition
Training Json Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb_20201014.json
Training Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb_20201014.log
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb_20201014-c9cdc656.pth
- Config: configs/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_rgb.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet50
Batch Size: 8
Epochs: 256
FLOPs: 54860480512
Parameters: 32454096
Pretrained: None
Resolution: short-side 256
Training Data: Kinetics-400
Training Resources: 32 GPUs
Modality: RGB
Name: slowonly_r50_8x8x1_256e_kinetics400_rgb
Results:
- Dataset: Kinetics-400
Metrics:
Top 1 Accuracy: 74.42
Top 5 Accuracy: 91.49
Task: Action Recognition
Training Json Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_8x8x1_256e_kinetics400_rgb/20200817_003320.log.json
Training Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_8x8x1_256e_kinetics400_rgb/20200817_003320.log
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_8x8x1_256e_kinetics400_rgb/slowonly_r50_256p_8x8x1_256e_kinetics400_rgb_20200820-75851a7d.pth
- Config: configs/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet50
Batch Size: 8
Epochs: 256
FLOPs: 27430649856
Parameters: 32454096
Pretrained: None
Resolution: short-side 320
Training Data: Kinetics-400
Training Resources: 16 GPUs
Modality: RGB
Name: slowonly_r50_4x16x1_256e_kinetics400_rgb
Results:
- Dataset: Kinetics-400
Metrics:
Top 1 Accuracy: 73.02
Top 5 Accuracy: 90.77
Task: Action Recognition
Training Json Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/slowonly_r50_4x16_73.02_90.77.log.json
Training Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/so_4x16.log
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/slowonly_r50_4x16x1_256e_kinetics400_rgb_20200704-a69556c6.pth
- Config: configs/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_rgb.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet50
Batch Size: 8
Epochs: 256
FLOPs: 54860480512
Parameters: 32454096
Pretrained: None
Resolution: short-side 320
Training Data: Kinetics-400
Training Resources: 24 GPUs
Modality: RGB
Name: slowonly_r50_8x8x1_256e_kinetics400_rgb
Results:
- Dataset: Kinetics-400
Metrics:
Top 1 Accuracy: 74.93
Top 5 Accuracy: 91.92
Task: Action Recognition
Training Json Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_rgb/slowonly_r50_8x8_74.93_91.92.log.json
Training Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_rgb/so_8x8.log
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_rgb/slowonly_r50_8x8x1_256e_kinetics400_rgb_20200703-a79c555a.pth
- Config: configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet50
Batch Size: 8
Epochs: 150
FLOPs: 27430649856
Parameters: 32454096
Pretrained: ImageNet
Resolution: short-side 320
Training Data: Kinetics-400
Training Resources: 16 GPUs
Modality: RGB
Name: slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb
Results:
- Dataset: Kinetics-400
Metrics:
Top 1 Accuracy: 73.39
Top 5 Accuracy: 91.12
Task: Action Recognition
Training Json Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb/slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb_20200912.json
Training Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb/slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb_20200912.log
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb/slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb_20200912-1e8fc736.pth
- Config: configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet50
Batch Size: 8
Epochs: 150
FLOPs: 54860480512
Parameters: 32454096
Pretrained: ImageNet
Resolution: short-side 320
Training Data: Kinetics-400
Training Resources: 32 GPUs
Modality: RGB
Name: slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb
Results:
- Dataset: Kinetics-400
Metrics:
Top 1 Accuracy: 75.55
Top 5 Accuracy: 92.04
Task: Action Recognition
Training Json Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb/slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb_20200912.json
Training Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb/slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb_20200912.log
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb/slowonly_imagenet_pretrained_r50_8x8x1_150e_kinetics400_rgb_20200912-3f9ce182.pth
- Config: configs/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_4x16x1_150e_kinetics400_rgb.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet50
Batch Size: 8
Epochs: 150
FLOPs: 38201098240
Parameters: 39808464
Pretrained: ImageNet
Resolution: short-side 320
Training Data: Kinetics-400
Training Resources: 16 GPUs
Modality: RGB
Name: slowonly_nl_embedded_gaussian_r50_4x16x1_150e_kinetics400_rgb
Results:
- Dataset: Kinetics-400
Metrics:
Top 1 Accuracy: 74.54
Top 5 Accuracy: 91.73
Task: Action Recognition
Training Json Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_4x16x1_150e_kinetics400_rgb/20210305_152630.log.json
Training Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_4x16x1_150e_kinetics400_rgb/20210305_152630.log
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_4x16x1_150e_kinetics400_rgb/slowonly_nl_embedded_gaussian_r50_4x16x1_150e_kinetics400_rgb_20210308-0d6e5a69.pth
- Config: configs/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_8x8x1_150e_kinetics400_rgb.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet50
Batch Size: 8
Epochs: 150
FLOPs: 76401377280
Parameters: 39808464
Pretrained: ImageNet
Resolution: short-side 320
Training Data: Kinetics-400
Training Resources: 32 GPUs
Modality: RGB
Name: slowonly_nl_embedded_gaussian_r50_8x8x1_150e_kinetics400_rgb
Results:
- Dataset: Kinetics-400
Metrics:
Top 1 Accuracy: 76.07
Top 5 Accuracy: 92.42
Task: Action Recognition
Training Json Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_8x8x1_150e_kinetics400_rgb/20210308_212250.log.json
Training Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_8x8x1_150e_kinetics400_rgb/20210308_212250.log
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_nl_embedded_gaussian_r50_8x8x1_150e_kinetics400_rgb/slowonly_nl_embedded_gaussian_r50_8x8x1_150e_kinetics400_rgb_20210308-e8dd9e82.pth
- Config: configs/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_flow.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet50
Batch Size: 24
Epochs: 256
FLOPs: 27225128960
Parameters: 32450960
Pretrained: ImageNet
Resolution: short-side 320
Training Data: Kinetics-400
Training Resources: 16 GPUs
Modality: Flow
Name: slowonly_r50_4x16x1_256e_kinetics400_flow
Results:
- Dataset: Kinetics-400
Metrics:
Top 1 Accuracy: 61.79
Top 5 Accuracy: 83.62
Task: Action Recognition
Training Json Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_flow/slowonly_r50_4x16x1_256e_kinetics400_flow_61.8_83.6.log.json
Training Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_flow/slowonly_r50_4x16x1_256e_kinetics400_flow_61.8_83.6.log
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_flow/slowonly_r50_4x16x1_256e_kinetics400_flow_20200704-decb8568.pth
- Config: configs/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_flow.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet50
Batch Size: 12
Epochs: 196
FLOPs: 54449438720
Parameters: 32450960
Pretrained: ImageNet
Resolution: short-side 320
Training Data: Kinetics-400
Training Resources: 32 GPUs
Modality: Flow
Name: slowonly_r50_8x8x1_196e_kinetics400_flow
Results:
- Dataset: Kinetics-400
Metrics:
Top 1 Accuracy: 65.76
Top 5 Accuracy: 86.25
Task: Action Recognition
Training Json Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_flow/slowonly_r50_8x8x1_196e_kinetics400_flow_65.8_86.3.log.json
Training Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_flow/slowonly_r50_8x8x1_196e_kinetics400_flow_65.8_86.3.log
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_flow/slowonly_r50_8x8x1_256e_kinetics400_flow_20200704-6b384243.pth
- Config: configs/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics600_rgb.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet50
Batch Size: 12
Epochs: 256
FLOPs: 54860890112
Parameters: 32863896
Pretrained: None
Resolution: short-side 256
Training Data: Kinetics-600
Training Resources: 32 GPUs
Modality: RGB
Name: slowonly_r50_video_8x8x1_256e_kinetics600_rgb
Results:
- Dataset: Kinetics-600
Metrics:
Top 1 Accuracy: 77.5
Top 5 Accuracy: 93.7
Task: Action Recognition
Training Json Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics600_rgb/slowonly_r50_video_8x8x1_256e_kinetics600_rgb_20201015.json
Training Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics600_rgb/slowonly_r50_video_8x8x1_256e_kinetics600_rgb_20201015.log
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics600_rgb/slowonly_r50_video_8x8x1_256e_kinetics600_rgb_20201015-81e5153e.pth
- Config: configs/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics700_rgb.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet50
Batch Size: 12
Epochs: 256
FLOPs: 54861094912
Parameters: 33068796
Pretrained: None
Resolution: short-side 256
Training Data: Kinetics-700
Training Resources: 32 GPUs
Modality: RGB
Name: slowonly_r50_video_8x8x1_256e_kinetics700_rgb
Results:
- Dataset: Kinetics-700
Metrics:
Top 1 Accuracy: 65.0
Top 5 Accuracy: 86.1
Task: Action Recognition
Training Json Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics700_rgb/slowonly_r50_video_8x8x1_256e_kinetics700_rgb_20201015.json
Training Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics700_rgb/slowonly_r50_video_8x8x1_256e_kinetics700_rgb_20201015.log
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_8x8x1_256e_kinetics700_rgb/slowonly_r50_video_8x8x1_256e_kinetics700_rgb_20201015-9250f662.pth
- Config: configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet50
Batch Size: 24
Epochs: 120
FLOPs: 27430649856
Parameters: 32454096
Pretrained: ImageNet
Resolution: short-side 256
Training Data: GYM99
Training Resources: 16 GPUs
Modality: RGB
Name: slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb
Results:
- Dataset: GYM99
Metrics:
Top 1 Accuracy: 79.3
mean Top 1 Accuracy: 70.2
Task: Action Recognition
Training Json Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb_20201111.json
Training Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb_20201111.log
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb_20201111-a9c34b54.pth
- Config: configs/recognition/slowonly/slowonly_kinetics_pretrained_r50_4x16x1_120e_gym99_flow.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet50
Batch Size: 24
Epochs: 120
FLOPs: 27225128960
Parameters: 32450960
Pretrained: Kinetics
Resolution: short-side 256
Training Data: GYM99
Training Resources: 16 GPUs
Modality: Flow
Name: slowonly_kinetics_pretrained_r50_4x16x1_120e_gym99_flow
Results:
- Dataset: GYM99
Metrics:
Top 1 Accuracy: 80.3
mean Top 1 Accuracy: 71.0
Task: Action Recognition
Training Json Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_kinetics_pretrained_r50_4x16x1_120e_gym99_flow/slowonly_kinetics_pretrained_r50_4x16x1_120e_gym99_flow_20201111.json
Training Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_kinetics_pretrained_r50_4x16x1_120e_gym99_flow/slowonly_kinetics_pretrained_r50_4x16x1_120e_gym99_flow_20201111.log
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_kinetics_pretrained_r50_4x16x1_120e_gym99_flow/slowonly_kinetics_pretrained_r50_4x16x1_120e_gym99_flow_20201111-66ecdb3c.pth
- Config: configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet50
Batch Size: 8
Epochs: 64
FLOPs: 54859716608
Parameters: 31689819
Pretrained: ImageNet
Resolution: height 100
Training Data: Jester
Training Resources: 8 GPUs
Modality: RGB
Name: slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb
Results:
- Dataset: Jester
Metrics:
Top 1 Accuracy: 97.2
Task: Action Recognition
Training Json Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb/slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb.log.json
Training Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb/slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb.log
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb/slowonly_imagenet_pretrained_r50_8x8x1_64e_jester_rgb-b56a5389.pth
- Config: configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_hmdb51_rgb.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet50
Batch Size: 8
Epochs: 64
FLOPs: 54859765760
Parameters: 31738995
Pretrained: ImageNet
Training Data: HMDB51
Training Resources: 8 GPUs
Modality: RGB
Name: slowonly_imagenet_pretrained_r50_8x4x1_64e_hmdb51_rgb
Results:
- Dataset: HMDB51
Metrics:
Top 1 Accuracy: 37.52
Top 5 Accuracy: 71.5
Task: Action Recognition
Training Json Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_hmdb51_rgb/20210605_185256.log.json
Training Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_hmdb51_rgb/20210605_185256.log
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_hmdb51_rgb/slowonly_imagenet_pretrained_r50_8x4x1_64e_hmdb51_rgb_20210630-16faeb6a.pth
- Config: configs/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_hmdb51_rgb.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet50
Batch Size: 8
Epochs: 40
FLOPs: 54859765760
Parameters: 31738995
Pretrained: Kinetics400
Training Data: HMDB51
Training Resources: 8 GPUs
Modality: RGB
Name: slowonly_imagenet_pretrained_r50_8x4x1_64e_hmdb51_rgb
Results:
- Dataset: HMDB51
Metrics:
Top 1 Accuracy: 65.95
Top 5 Accuracy: 91.05
Task: Action Recognition
Training Json Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_hmdb51_rgb/20210606_010153.log.json
Training Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_hmdb51_rgb/20210606_010153.log
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_hmdb51_rgb/slowonly_k400_pretrained_r50_8x4x1_40e_hmdb51_rgb_20210630-cee5f725.pth
- Config: configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_ucf101_rgb.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet50
Batch Size: 8
Epochs: 64
FLOPs: 54859868160
Parameters: 31841445
Pretrained: ImageNet
Training Data: UCF101
Training Resources: 8 GPUs
Modality: RGB
Name: slowonly_imagenet_pretrained_r50_8x4x1_64e_ucf101_rgb
Results:
- Dataset: UCF101
Metrics:
Top 1 Accuracy: 71.35
Top 5 Accuracy: 89.35
Task: Action Recognition
Training Json Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_ucf101_rgb/20210605_213503.log.json
Training Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_ucf101_rgb/20210605_213503.log
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_ucf101_rgb/slowonly_imagenet_pretrained_r50_8x4x1_64e_ucf101_rgb_20210630-181e1661.pth
- Config: configs/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_ucf101_rgb.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet50
Batch Size: 8
Epochs: 40
FLOPs: 54859868160
Parameters: 31841445
Pretrained: Kinetics400
Training Data: UCF101
Training Resources: 8 GPUs
Modality: RGB
Name: slowonly_k400_pretrained_r50_8x4x1_40e_ucf101_rgb
Results:
- Dataset: UCF101
Metrics:
Top 1 Accuracy: 92.78
Top 5 Accuracy: 99.42
Task: Action Recognition
Training Json Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_ucf101_rgb/20210606_010231.log.json
Training Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_ucf101_rgb/20210606_010231.log
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_k400_pretrained_r50_8x4x1_40e_ucf101_rgb/slowonly_k400_pretrained_r50_8x4x1_40e_ucf101_rgb_20210630-ee8c850f.pth
- Config: configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb.py
In Collection: SlowOnly
Metadata:
Architecture: ResNet50
Batch Size: 8
Epochs: 64
FLOPs: 53907910656
Parameters: 31991022
Pretrained: ImageNet
Training Data: SthV1
Training Resources: 8 GPUs
Modality: RGB
Name: slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb
Results:
- Dataset: SthV1
Metrics:
Top 1 Accuracy: 47.76
Top 5 Accuracy: 77.49
Task: Action Recognition
Training Json Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb/slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb.json
Training Log: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb/slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb.log
Weights: https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb/slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb_20211202-d034ff12.pth
configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb.py 0 → 100644
View file @ 5b3e36dc
_base_ = [
'../../_base_/models/slowonly_r50.py', '../../_base_/default_runtime.py'
]
# dataset settings
dataset_type = 'RawframeDataset'
data_root = 'data/gym/rawframes'
data_root_val = 'data/gym/rawframes'
ann_file_train = 'data/gym/annotations/gym99_train_list_rawframes.txt'
ann_file_val = 'data/gym/annotations/gym99_val_list_rawframes.txt'
ann_file_test = 'data/gym/annotations/gym99_val_list_rawframes.txt'
img_norm_cfg = dict(
mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_bgr=False)
train_pipeline = [
dict(type='SampleFrames', clip_len=4, frame_interval=16, num_clips=1),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='RandomResizedCrop'),
dict(type='Resize', scale=(224, 224), keep_ratio=False),
dict(type='Flip', flip_ratio=0.5),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs', 'label'])
]
val_pipeline = [
dict(
type='SampleFrames',
clip_len=4,
frame_interval=16,
num_clips=1,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='CenterCrop', crop_size=224),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
test_pipeline = [
dict(
type='SampleFrames',
clip_len=4,
frame_interval=16,
num_clips=10,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='ThreeCrop', crop_size=256),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
data = dict(
videos_per_gpu=24,
workers_per_gpu=2,
test_dataloader=dict(videos_per_gpu=1),
train=dict(
type=dataset_type,
ann_file=ann_file_train,
data_prefix=data_root,
pipeline=train_pipeline),
val=dict(
type=dataset_type,
ann_file=ann_file_val,
data_prefix=data_root_val,
pipeline=val_pipeline),
test=dict(
type=dataset_type,
ann_file=ann_file_test,
data_prefix=data_root_val,
pipeline=test_pipeline))
evaluation = dict(
interval=5, metrics=['top_k_accuracy', 'mean_class_accuracy'])
# optimizer
optimizer = dict(
type='SGD', lr=0.03, momentum=0.9,
weight_decay=0.0001) # this lr is used for 8 gpus
optimizer_config = dict(grad_clip=dict(max_norm=40, norm_type=2))
# learning policy
lr_config = dict(policy='step', step=[90, 110])
total_epochs = 120
# runtime settings
work_dir = './work_dirs/slowonly_imagenet_pretrained_r50_4x16x1_120e_gym99_rgb'
find_unused_parameters = False
configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_4x16x1_150e_kinetics400_rgb.py 0 → 100644
View file @ 5b3e36dc
_base_ = [
'../../_base_/models/slowonly_r50.py', '../../_base_/default_runtime.py'
]
# dataset settings
dataset_type = 'RawframeDataset'
data_root = 'data/kinetics400/rawframes_train'
data_root_val = 'data/kinetics400/rawframes_val'
ann_file_train = 'data/kinetics400/kinetics400_train_list_rawframes.txt'
ann_file_val = 'data/kinetics400/kinetics400_val_list_rawframes.txt'
ann_file_test = 'data/kinetics400/kinetics400_val_list_rawframes.txt'
img_norm_cfg = dict(
mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_bgr=False)
train_pipeline = [
dict(type='SampleFrames', clip_len=4, frame_interval=16, num_clips=1),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='RandomResizedCrop'),
dict(type='Resize', scale=(224, 224), keep_ratio=False),
dict(type='Flip', flip_ratio=0.5),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs', 'label'])
]
val_pipeline = [
dict(
type='SampleFrames',
clip_len=4,
frame_interval=16,
num_clips=1,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='CenterCrop', crop_size=224),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
test_pipeline = [
dict(
type='SampleFrames',
clip_len=4,
frame_interval=16,
num_clips=10,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='ThreeCrop', crop_size=256),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
data = dict(
videos_per_gpu=8,
workers_per_gpu=2,
test_dataloader=dict(videos_per_gpu=1),
train=dict(
type=dataset_type,
ann_file=ann_file_train,
data_prefix=data_root,
pipeline=train_pipeline),
val=dict(
type=dataset_type,
ann_file=ann_file_val,
data_prefix=data_root_val,
pipeline=val_pipeline),
test=dict(
type=dataset_type,
ann_file=ann_file_test,
data_prefix=data_root_val,
pipeline=test_pipeline))
evaluation = dict(
interval=5, metrics=['top_k_accuracy', 'mean_class_accuracy'])
# optimizer
optimizer = dict(
type='SGD', lr=0.01, momentum=0.9,
weight_decay=0.0001) # this lr is used for 8 gpus
optimizer_config = dict(grad_clip=dict(max_norm=40, norm_type=2))
# learning policy
lr_config = dict(
policy='step',
step=[90, 130],
warmup='linear',
warmup_by_epoch=True,
warmup_iters=10)
total_epochs = 150
# runtime settings
checkpoint_config = dict(interval=4)
work_dir = ('./work_dirs/slowonly_imagenet_pretrained_r50_4x16x1_150e'
'_kinetics400_rgb')
find_unused_parameters = False
configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_hmdb51_rgb.py 0 → 100644
View file @ 5b3e36dc
_base_ = [
'../../_base_/models/slowonly_r50.py',
'../../_base_/schedules/sgd_150e_warmup.py',
'../../_base_/default_runtime.py'
]
# model settings
model = dict(cls_head=dict(num_classes=51))
# dataset settings
split = 1
dataset_type = 'RawframeDataset'
data_root = 'data/hmdb51/rawframes'
data_root_val = 'data/hmdb51/rawframes'
ann_file_train = f'data/hmdb51/hmdb51_train_split_{split}_rawframes.txt'
ann_file_val = f'data/hmdb51/hmdb51_val_split_{split}_rawframes.txt'
ann_file_test = f'data/hmdb51/hmdb51_val_split_{split}_rawframes.txt'
img_norm_cfg = dict(
mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_bgr=False)
train_pipeline = [
dict(type='SampleFrames', clip_len=8, frame_interval=4, num_clips=1),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='RandomResizedCrop'),
dict(type='Resize', scale=(224, 224), keep_ratio=False),
dict(type='Flip', flip_ratio=0.5),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs', 'label'])
]
val_pipeline = [
dict(
type='SampleFrames',
clip_len=8,
frame_interval=4,
num_clips=1,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='CenterCrop', crop_size=224),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
test_pipeline = [
dict(
type='SampleFrames',
clip_len=8,
frame_interval=4,
num_clips=10,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='ThreeCrop', crop_size=256),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
data = dict(
videos_per_gpu=8,
workers_per_gpu=2,
test_dataloader=dict(videos_per_gpu=1),
train=dict(
type=dataset_type,
ann_file=ann_file_train,
data_prefix=data_root,
pipeline=train_pipeline),
val=dict(
type=dataset_type,
ann_file=ann_file_val,
data_prefix=data_root_val,
pipeline=val_pipeline),
test=dict(
type=dataset_type,
ann_file=ann_file_test,
data_prefix=data_root_val,
pipeline=test_pipeline))
evaluation = dict(
interval=1, metrics=['top_k_accuracy', 'mean_class_accuracy'])
# optimizer
optimizer = dict(lr=0.1) # this lr is used for 8 gpus
# learning policy
lr_config = dict(policy='CosineAnnealing', min_lr=0, by_epoch=False)
total_epochs = 64
# runtime settings
work_dir = './work_dirs/slowonly_r50_8x4x1_64e_hmdb51_rgb'
configs/recognition/slowonly/slowonly_imagenet_pretrained_r50_8x4x1_64e_sthv1_rgb.py 0 → 100644
View file @ 5b3e36dc
_base_ = [
'../../_base_/models/slowonly_r50.py',
'../../_base_/schedules/sgd_150e_warmup.py',
'../../_base_/default_runtime.py'
]
# model settings
model = dict(backbone=dict(with_pool1=False), cls_head=dict(num_classes=174))
# dataset settings
dataset_type = 'RawframeDataset'
data_root = 'data/sthv1/rawframes'
data_root_val = 'data/sthv1/rawframes'
ann_file_train = 'data/sthv1/sthv1_train_list_rawframes.txt'
ann_file_val = 'data/sthv1/sthv1_val_list_rawframes.txt'
ann_file_test = 'data/sthv1/sthv1_val_list_rawframes.txt'
img_norm_cfg = dict(
mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_bgr=False)
train_pipeline = [
dict(type='SampleFrames', clip_len=8, frame_interval=4, num_clips=1),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 128)),
dict(type='RandomResizedCrop'),
dict(type='Resize', scale=(112, 112), keep_ratio=False),
dict(type='Flip', flip_ratio=0.5),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs', 'label'])
]
val_pipeline = [
dict(
type='SampleFrames',
clip_len=8,
frame_interval=4,
num_clips=1,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 128)),
dict(type='CenterCrop', crop_size=112),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
test_pipeline = [
dict(
type='SampleFrames',
clip_len=8,
frame_interval=4,
num_clips=10,
test_mode=True),
dict(type='RawFrameDecode'),
dict(type='Resize', scale=(-1, 128)),
dict(type='ThreeCrop', crop_size=128),
dict(type='Normalize', **img_norm_cfg),
dict(type='FormatShape', input_format='NCTHW'),
dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
dict(type='ToTensor', keys=['imgs'])
]
data = dict(
videos_per_gpu=8,
workers_per_gpu=2,
test_dataloader=dict(videos_per_gpu=1),
train=dict(
type=dataset_type,
ann_file=ann_file_train,
filename_tmpl='{:05}.jpg',
data_prefix=data_root,
pipeline=train_pipeline),
val=dict(
type=dataset_type,
ann_file=ann_file_val,
filename_tmpl='{:05}.jpg',
data_prefix=data_root_val,
pipeline=val_pipeline),
test=dict(
type=dataset_type,
ann_file=ann_file_test,
filename_tmpl='{:05}.jpg',
data_prefix=data_root_val,
pipeline=test_pipeline))
evaluation = dict(
interval=1, metrics=['top_k_accuracy', 'mean_class_accuracy'])
# optimizer
optimizer = dict(lr=0.1) # this lr is used for 8 gpus
# learning policy
lr_config = dict(
policy='CosineAnnealing',
min_lr=0,
warmup='linear',
warmup_by_epoch=True,
warmup_iters=10)
total_epochs = 64
# runtime settings
work_dir = './work_dirs/slowonly_r50_8x4x1_64e_sthv1_rgb'
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