Skip to content

GitLab

Unverified Commit dc6de4df authored by Zhe Chen's avatar Zhe Chen Committed by GitHub
Browse files

Fix bug in classification & detection (#252) 

* fix drop_path_rate in classification

* add ms in detection 3x config
parent 3e083be9
Show whitespace changes
Inline Side-by-side
Showing with 133 additions and 3 deletions
classification/models/build.py
View file @ dc6de4df
......@@ -21,6 +21,7 @@ def build_model(config):
post_norm=config.MODEL.INTERN_IMAGE.POST_NORM,
mlp_ratio=config.MODEL.INTERN_IMAGE.MLP_RATIO,
with_cp=config.TRAIN.USE_CHECKPOINT,
drop_path_rate=config.MODEL.DROP_PATH_RATE,
res_post_norm=config.MODEL.INTERN_IMAGE.RES_POST_NORM, # for InternImage-H/G
dw_kernel_size=config.MODEL.INTERN_IMAGE.DW_KERNEL_SIZE, # for InternImage-H/G
use_clip_projector=config.MODEL.INTERN_IMAGE.USE_CLIP_PROJECTOR, # for InternImage-H/G
......
detection/configs/coco/mask_rcnn_internimage_b_fpn_3x_coco.py
View file @ dc6de4df
......@@ -32,8 +32,51 @@ model = dict(
in_channels=[112, 224, 448, 896],
out_channels=256,
num_outs=5))
img_norm_cfg = dict(
mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
# augmentation strategy originates from DETR / Sparse RCNN
train_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='LoadAnnotations', with_bbox=True, with_mask=True),
dict(type='RandomFlip', flip_ratio=0.5),
dict(type='AutoAugment',
policies=[
[
dict(type='Resize',
img_scale=[(480, 1333), (512, 1333), (544, 1333), (576, 1333),
(608, 1333), (640, 1333), (672, 1333), (704, 1333),
(736, 1333), (768, 1333), (800, 1333)],
multiscale_mode='value',
keep_ratio=True)
],
[
dict(type='Resize',
img_scale=[(400, 1333), (500, 1333), (600, 1333)],
multiscale_mode='value',
keep_ratio=True),
dict(type='RandomCrop',
crop_type='absolute_range',
crop_size=(384, 600),
allow_negative_crop=True),
dict(type='Resize',
img_scale=[(480, 1333), (512, 1333), (544, 1333),
(576, 1333), (608, 1333), (640, 1333),
(672, 1333), (704, 1333), (736, 1333),
(768, 1333), (800, 1333)],
multiscale_mode='value',
override=True,
keep_ratio=True)
]
]),
dict(type='Normalize', **img_norm_cfg),
dict(type='Pad', size_divisor=32),
dict(type='DefaultFormatBundle'),
dict(type='Collect', keys=['img', 'gt_bboxes', 'gt_labels', 'gt_masks']),
]
# By default, models are trained on 8 GPUs with 2 images per GPU
data = dict(samples_per_gpu=2)
data = dict(
samples_per_gpu=2,
train=dict(pipeline=train_pipeline))
optimizer = dict(
_delete_=True, type='AdamW', lr=0.0001, weight_decay=0.05,
constructor='CustomLayerDecayOptimizerConstructor',
......
detection/configs/coco/mask_rcnn_internimage_s_fpn_3x_coco.py
View file @ dc6de4df
......@@ -32,8 +32,51 @@ model = dict(
in_channels=[80, 160, 320, 640],
out_channels=256,
num_outs=5))
img_norm_cfg = dict(
mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
# augmentation strategy originates from DETR / Sparse RCNN
train_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='LoadAnnotations', with_bbox=True, with_mask=True),
dict(type='RandomFlip', flip_ratio=0.5),
dict(type='AutoAugment',
policies=[
[
dict(type='Resize',
img_scale=[(480, 1333), (512, 1333), (544, 1333), (576, 1333),
(608, 1333), (640, 1333), (672, 1333), (704, 1333),
(736, 1333), (768, 1333), (800, 1333)],
multiscale_mode='value',
keep_ratio=True)
],
[
dict(type='Resize',
img_scale=[(400, 1333), (500, 1333), (600, 1333)],
multiscale_mode='value',
keep_ratio=True),
dict(type='RandomCrop',
crop_type='absolute_range',
crop_size=(384, 600),
allow_negative_crop=True),
dict(type='Resize',
img_scale=[(480, 1333), (512, 1333), (544, 1333),
(576, 1333), (608, 1333), (640, 1333),
(672, 1333), (704, 1333), (736, 1333),
(768, 1333), (800, 1333)],
multiscale_mode='value',
override=True,
keep_ratio=True)
]
]),
dict(type='Normalize', **img_norm_cfg),
dict(type='Pad', size_divisor=32),
dict(type='DefaultFormatBundle'),
dict(type='Collect', keys=['img', 'gt_bboxes', 'gt_labels', 'gt_masks']),
]
# By default, models are trained on 8 GPUs with 2 images per GPU
data = dict(samples_per_gpu=2)
data = dict(
samples_per_gpu=2,
train=dict(pipeline=train_pipeline))
optimizer = dict(
_delete_=True, type='AdamW', lr=0.0001, weight_decay=0.05,
constructor='CustomLayerDecayOptimizerConstructor',
......
detection/configs/coco/mask_rcnn_internimage_t_fpn_3x_coco.py
View file @ dc6de4df
......@@ -32,8 +32,51 @@ model = dict(
in_channels=[64, 128, 256, 512],
out_channels=256,
num_outs=5))
img_norm_cfg = dict(
mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
# augmentation strategy originates from DETR / Sparse RCNN
train_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='LoadAnnotations', with_bbox=True, with_mask=True),
dict(type='RandomFlip', flip_ratio=0.5),
dict(type='AutoAugment',
policies=[
[
dict(type='Resize',
img_scale=[(480, 1333), (512, 1333), (544, 1333), (576, 1333),
(608, 1333), (640, 1333), (672, 1333), (704, 1333),
(736, 1333), (768, 1333), (800, 1333)],
multiscale_mode='value',
keep_ratio=True)
],
[
dict(type='Resize',
img_scale=[(400, 1333), (500, 1333), (600, 1333)],
multiscale_mode='value',
keep_ratio=True),
dict(type='RandomCrop',
crop_type='absolute_range',
crop_size=(384, 600),
allow_negative_crop=True),
dict(type='Resize',
img_scale=[(480, 1333), (512, 1333), (544, 1333),
(576, 1333), (608, 1333), (640, 1333),
(672, 1333), (704, 1333), (736, 1333),
(768, 1333), (800, 1333)],
multiscale_mode='value',
override=True,
keep_ratio=True)
]
]),
dict(type='Normalize', **img_norm_cfg),
dict(type='Pad', size_divisor=32),
dict(type='DefaultFormatBundle'),
dict(type='Collect', keys=['img', 'gt_bboxes', 'gt_labels', 'gt_masks']),
]
# By default, models are trained on 8 GPUs with 2 images per GPU
data = dict(samples_per_gpu=2)
data = dict(
samples_per_gpu=2,
train=dict(pipeline=train_pipeline))
optimizer = dict(
_delete_=True, type='AdamW', lr=0.0001, weight_decay=0.05,
constructor='CustomLayerDecayOptimizerConstructor',
......
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