Skip to content

GitLab

Commit 37c8cebc authored by Sugon_ldc's avatar Sugon_ldc
Browse files

add new model

parents
Pipeline #318 failed with stages
in 0 seconds
Hide whitespace changes
Inline Side-by-side
Showing with 2309 additions and 0 deletions
configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnetv1d50_coco_384x288.py 0 → 100755
View file @ 37c8cebc
_base_ = [
'../../../../_base_/default_runtime.py',
'../../../../_base_/datasets/coco.py'
]
evaluation = dict(interval=10, metric='mAP', save_best='AP')
optimizer = dict(
type='Adam',
lr=5e-4,
)
optimizer_config = dict(grad_clip=None)
# learning policy
lr_config = dict(
policy='step',
warmup='linear',
warmup_iters=500,
warmup_ratio=0.001,
step=[170, 200])
total_epochs = 210
channel_cfg = dict(
num_output_channels=17,
dataset_joints=17,
dataset_channel=[
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16],
],
inference_channel=[
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16
])
# model settings
model = dict(
type='TopDown',
pretrained='mmcls://resnet50_v1d',
backbone=dict(type='ResNetV1d', depth=50),
keypoint_head=dict(
type='TopdownHeatmapSimpleHead',
in_channels=2048,
out_channels=channel_cfg['num_output_channels'],
loss_keypoint=dict(type='JointsMSELoss', use_target_weight=True)),
train_cfg=dict(),
test_cfg=dict(
flip_test=True,
post_process='default',
shift_heatmap=True,
modulate_kernel=11))
data_cfg = dict(
image_size=[288, 384],
heatmap_size=[72, 96],
num_output_channels=channel_cfg['num_output_channels'],
num_joints=channel_cfg['dataset_joints'],
dataset_channel=channel_cfg['dataset_channel'],
inference_channel=channel_cfg['inference_channel'],
soft_nms=False,
nms_thr=1.0,
oks_thr=0.9,
vis_thr=0.2,
use_gt_bbox=False,
det_bbox_thr=0.0,
bbox_file='data/coco/person_detection_results/'
'COCO_val2017_detections_AP_H_56_person.json',
)
train_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownRandomShiftBboxCenter', shift_factor=0.16, prob=0.3),
dict(type='TopDownRandomFlip', flip_prob=0.5),
dict(
type='TopDownHalfBodyTransform',
num_joints_half_body=8,
prob_half_body=0.3),
dict(
type='TopDownGetRandomScaleRotation', rot_factor=40, scale_factor=0.5),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(type='TopDownGenerateTarget', sigma=3),
dict(
type='Collect',
keys=['img', 'target', 'target_weight'],
meta_keys=[
'image_file', 'joints_3d', 'joints_3d_visible', 'center', 'scale',
'rotation', 'bbox_score', 'flip_pairs'
]),
]
val_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(
type='Collect',
keys=['img'],
meta_keys=[
'image_file', 'center', 'scale', 'rotation', 'bbox_score',
'flip_pairs'
]),
]
test_pipeline = val_pipeline
data_root = 'data/coco'
data = dict(
samples_per_gpu=64,
workers_per_gpu=2,
val_dataloader=dict(samples_per_gpu=32),
test_dataloader=dict(samples_per_gpu=32),
train=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_train2017.json',
img_prefix=f'{data_root}/train2017/',
data_cfg=data_cfg,
pipeline=train_pipeline,
dataset_info={{_base_.dataset_info}}),
val=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=val_pipeline,
dataset_info={{_base_.dataset_info}}),
test=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=test_pipeline,
dataset_info={{_base_.dataset_info}}),
)
configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnetv1d_coco.md 0 → 100755
View file @ 37c8cebc
<!-- [BACKBONE] -->
<details>
<summary align="right"><a href="http://openaccess.thecvf.com/content_CVPR_2019/html/He_Bag_of_Tricks_for_Image_Classification_with_Convolutional_Neural_Networks_CVPR_2019_paper.html">ResNetV1D (CVPR'2019)</a></summary>
```bibtex
@inproceedings{he2019bag,
title={Bag of tricks for image classification with convolutional neural networks},
author={He, Tong and Zhang, Zhi and Zhang, Hang and Zhang, Zhongyue and Xie, Junyuan and Li, Mu},
booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
pages={558--567},
year={2019}
}
```
</details>
<!-- [DATASET] -->
<details>
<summary align="right"><a href="https://link.springer.com/chapter/10.1007/978-3-319-10602-1_48">COCO (ECCV'2014)</a></summary>
```bibtex
@inproceedings{lin2014microsoft,
title={Microsoft coco: Common objects in context},
author={Lin, Tsung-Yi and Maire, Michael and Belongie, Serge and Hays, James and Perona, Pietro and Ramanan, Deva and Doll{\'a}r, Piotr and Zitnick, C Lawrence},
booktitle={European conference on computer vision},
pages={740--755},
year={2014},
organization={Springer}
}
```
</details>
Results on COCO val2017 with detector having human AP of 56.4 on COCO val2017 dataset
| Arch | Input Size | AP | AP<sup>50</sup> | AP<sup>75</sup> | AR | AR<sup>50</sup> | ckpt | log |
| :-------------------------------------------- | :--------: | :---: | :-------------: | :-------------: | :---: | :-------------: | :-------------------------------------------: | :-------------------------------------------: |
| [pose_resnetv1d_50](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnetv1d50_coco_256x192.py) | 256x192 | 0.722 | 0.897 | 0.799 | 0.777 | 0.933 | [ckpt](https://download.openmmlab.com/mmpose/top_down/resnetv1d/resnetv1d50_coco_256x192-a243b840_20200727.pth) | [log](https://download.openmmlab.com/mmpose/top_down/resnetv1d/resnetv1d50_coco_256x192_20200727.log.json) |
| [pose_resnetv1d_50](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnetv1d50_coco_384x288.py) | 384x288 | 0.730 | 0.900 | 0.799 | 0.780 | 0.934 | [ckpt](https://download.openmmlab.com/mmpose/top_down/resnetv1d/resnetv1d50_coco_384x288-01f3fbb9_20200727.pth) | [log](https://download.openmmlab.com/mmpose/top_down/resnetv1d/resnetv1d50_coco_384x288_20200727.log.json) |
| [pose_resnetv1d_101](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnetv1d101_coco_256x192.py) | 256x192 | 0.731 | 0.899 | 0.809 | 0.786 | 0.938 | [ckpt](https://download.openmmlab.com/mmpose/top_down/resnetv1d/resnetv1d101_coco_256x192-5bd08cab_20200727.pth) | [log](https://download.openmmlab.com/mmpose/top_down/resnetv1d/resnetv1d101_coco_256x192_20200727.log.json) |
| [pose_resnetv1d_101](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnetv1d101_coco_384x288.py) | 384x288 | 0.748 | 0.902 | 0.816 | 0.799 | 0.939 | [ckpt](https://download.openmmlab.com/mmpose/top_down/resnetv1d/resnetv1d101_coco_384x288-5f9e421d_20200730.pth) | [log](https://download.openmmlab.com/mmpose/top_down/resnetv1d/resnetv1d101_coco_384x288-20200730.log.json) |
| [pose_resnetv1d_152](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnetv1d152_coco_256x192.py) | 256x192 | 0.737 | 0.902 | 0.812 | 0.791 | 0.940 | [ckpt](https://download.openmmlab.com/mmpose/top_down/resnetv1d/resnetv1d152_coco_256x192-c4df51dc_20200727.pth) | [log](https://download.openmmlab.com/mmpose/top_down/resnetv1d/resnetv1d152_coco_256x192_20200727.log.json) |
| [pose_resnetv1d_152](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnetv1d152_coco_384x288.py) | 384x288 | 0.752 | 0.909 | 0.821 | 0.802 | 0.944 | [ckpt](https://download.openmmlab.com/mmpose/top_down/resnetv1d/resnetv1d152_coco_384x288-626c622d_20200730.pth) | [log](https://download.openmmlab.com/mmpose/top_down/resnetv1d/resnetv1d152_coco_384x288-20200730.log.json) |
configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnetv1d_coco.yml 0 → 100755
View file @ 37c8cebc
Collections:
- Name: ResNetV1D
Paper:
Title: Bag of tricks for image classification with convolutional neural networks
URL: http://openaccess.thecvf.com/content_CVPR_2019/html/He_Bag_of_Tricks_for_Image_Classification_with_Convolutional_Neural_Networks_CVPR_2019_paper.html
README: https://github.com/open-mmlab/mmpose/blob/master/docs/en/papers/backbones/resnetv1d.md
Models:
- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnetv1d50_coco_256x192.py
In Collection: ResNetV1D
Metadata:
Architecture: &id001
- ResNetV1D
Training Data: COCO
Name: topdown_heatmap_resnetv1d50_coco_256x192
Results:
- Dataset: COCO
Metrics:
AP: 0.722
AP@0.5: 0.897
AP@0.75: 0.799
AR: 0.777
AR@0.5: 0.933
Task: Body 2D Keypoint
Weights: https://download.openmmlab.com/mmpose/top_down/resnetv1d/resnetv1d50_coco_256x192-a243b840_20200727.pth
- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnetv1d50_coco_384x288.py
In Collection: ResNetV1D
Metadata:
Architecture: *id001
Training Data: COCO
Name: topdown_heatmap_resnetv1d50_coco_384x288
Results:
- Dataset: COCO
Metrics:
AP: 0.73
AP@0.5: 0.9
AP@0.75: 0.799
AR: 0.78
AR@0.5: 0.934
Task: Body 2D Keypoint
Weights: https://download.openmmlab.com/mmpose/top_down/resnetv1d/resnetv1d50_coco_384x288-01f3fbb9_20200727.pth
- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnetv1d101_coco_256x192.py
In Collection: ResNetV1D
Metadata:
Architecture: *id001
Training Data: COCO
Name: topdown_heatmap_resnetv1d101_coco_256x192
Results:
- Dataset: COCO
Metrics:
AP: 0.731
AP@0.5: 0.899
AP@0.75: 0.809
AR: 0.786
AR@0.5: 0.938
Task: Body 2D Keypoint
Weights: https://download.openmmlab.com/mmpose/top_down/resnetv1d/resnetv1d101_coco_256x192-5bd08cab_20200727.pth
- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnetv1d101_coco_384x288.py
In Collection: ResNetV1D
Metadata:
Architecture: *id001
Training Data: COCO
Name: topdown_heatmap_resnetv1d101_coco_384x288
Results:
- Dataset: COCO
Metrics:
AP: 0.748
AP@0.5: 0.902
AP@0.75: 0.816
AR: 0.799
AR@0.5: 0.939
Task: Body 2D Keypoint
Weights: https://download.openmmlab.com/mmpose/top_down/resnetv1d/resnetv1d101_coco_384x288-5f9e421d_20200730.pth
- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnetv1d152_coco_256x192.py
In Collection: ResNetV1D
Metadata:
Architecture: *id001
Training Data: COCO
Name: topdown_heatmap_resnetv1d152_coco_256x192
Results:
- Dataset: COCO
Metrics:
AP: 0.737
AP@0.5: 0.902
AP@0.75: 0.812
AR: 0.791
AR@0.5: 0.94
Task: Body 2D Keypoint
Weights: https://download.openmmlab.com/mmpose/top_down/resnetv1d/resnetv1d152_coco_256x192-c4df51dc_20200727.pth
- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnetv1d152_coco_384x288.py
In Collection: ResNetV1D
Metadata:
Architecture: *id001
Training Data: COCO
Name: topdown_heatmap_resnetv1d152_coco_384x288
Results:
- Dataset: COCO
Metrics:
AP: 0.752
AP@0.5: 0.909
AP@0.75: 0.821
AR: 0.802
AR@0.5: 0.944
Task: Body 2D Keypoint
Weights: https://download.openmmlab.com/mmpose/top_down/resnetv1d/resnetv1d152_coco_384x288-626c622d_20200730.pth
configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnext101_coco_256x192.py 0 → 100755
View file @ 37c8cebc
_base_ = [
'../../../../_base_/default_runtime.py',
'../../../../_base_/datasets/coco.py'
]
evaluation = dict(interval=10, metric='mAP', save_best='AP')
optimizer = dict(
type='Adam',
lr=5e-4,
)
optimizer_config = dict(grad_clip=None)
# learning policy
lr_config = dict(
policy='step',
warmup='linear',
warmup_iters=500,
warmup_ratio=0.001,
step=[170, 200])
total_epochs = 210
channel_cfg = dict(
num_output_channels=17,
dataset_joints=17,
dataset_channel=[
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16],
],
inference_channel=[
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16
])
# model settings
model = dict(
type='TopDown',
pretrained='mmcls://resnext101_32x4d',
backbone=dict(type='ResNeXt', depth=101),
keypoint_head=dict(
type='TopdownHeatmapSimpleHead',
in_channels=2048,
out_channels=channel_cfg['num_output_channels'],
loss_keypoint=dict(type='JointsMSELoss', use_target_weight=True)),
train_cfg=dict(),
test_cfg=dict(
flip_test=True,
post_process='default',
shift_heatmap=True,
modulate_kernel=11))
data_cfg = dict(
image_size=[192, 256],
heatmap_size=[48, 64],
num_output_channels=channel_cfg['num_output_channels'],
num_joints=channel_cfg['dataset_joints'],
dataset_channel=channel_cfg['dataset_channel'],
inference_channel=channel_cfg['inference_channel'],
soft_nms=False,
nms_thr=1.0,
oks_thr=0.9,
vis_thr=0.2,
use_gt_bbox=False,
det_bbox_thr=0.0,
bbox_file='data/coco/person_detection_results/'
'COCO_val2017_detections_AP_H_56_person.json',
)
train_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownRandomShiftBboxCenter', shift_factor=0.16, prob=0.3),
dict(type='TopDownRandomFlip', flip_prob=0.5),
dict(
type='TopDownHalfBodyTransform',
num_joints_half_body=8,
prob_half_body=0.3),
dict(
type='TopDownGetRandomScaleRotation', rot_factor=40, scale_factor=0.5),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(type='TopDownGenerateTarget', sigma=2),
dict(
type='Collect',
keys=['img', 'target', 'target_weight'],
meta_keys=[
'image_file', 'joints_3d', 'joints_3d_visible', 'center', 'scale',
'rotation', 'bbox_score', 'flip_pairs'
]),
]
val_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(
type='Collect',
keys=['img'],
meta_keys=[
'image_file', 'center', 'scale', 'rotation', 'bbox_score',
'flip_pairs'
]),
]
test_pipeline = val_pipeline
data_root = 'data/coco'
data = dict(
samples_per_gpu=64,
workers_per_gpu=2,
val_dataloader=dict(samples_per_gpu=32),
test_dataloader=dict(samples_per_gpu=32),
train=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_train2017.json',
img_prefix=f'{data_root}/train2017/',
data_cfg=data_cfg,
pipeline=train_pipeline,
dataset_info={{_base_.dataset_info}}),
val=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=val_pipeline,
dataset_info={{_base_.dataset_info}}),
test=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=test_pipeline,
dataset_info={{_base_.dataset_info}}),
)
configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnext101_coco_384x288.py 0 → 100755
View file @ 37c8cebc
_base_ = [
'../../../../_base_/default_runtime.py',
'../../../../_base_/datasets/coco.py'
]
evaluation = dict(interval=10, metric='mAP', save_best='AP')
optimizer = dict(
type='Adam',
lr=5e-4,
)
optimizer_config = dict(grad_clip=None)
# learning policy
lr_config = dict(
policy='step',
warmup='linear',
warmup_iters=500,
warmup_ratio=0.001,
step=[170, 200])
total_epochs = 210
channel_cfg = dict(
num_output_channels=17,
dataset_joints=17,
dataset_channel=[
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16],
],
inference_channel=[
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16
])
# model settings
model = dict(
type='TopDown',
pretrained='mmcls://resnext101_32x4d',
backbone=dict(type='ResNeXt', depth=101),
keypoint_head=dict(
type='TopdownHeatmapSimpleHead',
in_channels=2048,
out_channels=channel_cfg['num_output_channels'],
loss_keypoint=dict(type='JointsMSELoss', use_target_weight=True)),
train_cfg=dict(),
test_cfg=dict(
flip_test=True,
post_process='default',
shift_heatmap=True,
modulate_kernel=11))
data_cfg = dict(
image_size=[288, 384],
heatmap_size=[72, 96],
num_output_channels=channel_cfg['num_output_channels'],
num_joints=channel_cfg['dataset_joints'],
dataset_channel=channel_cfg['dataset_channel'],
inference_channel=channel_cfg['inference_channel'],
soft_nms=False,
nms_thr=1.0,
oks_thr=0.9,
vis_thr=0.2,
use_gt_bbox=False,
det_bbox_thr=0.0,
bbox_file='data/coco/person_detection_results/'
'COCO_val2017_detections_AP_H_56_person.json',
)
train_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownRandomShiftBboxCenter', shift_factor=0.16, prob=0.3),
dict(type='TopDownRandomFlip', flip_prob=0.5),
dict(
type='TopDownHalfBodyTransform',
num_joints_half_body=8,
prob_half_body=0.3),
dict(
type='TopDownGetRandomScaleRotation', rot_factor=40, scale_factor=0.5),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(type='TopDownGenerateTarget', sigma=3),
dict(
type='Collect',
keys=['img', 'target', 'target_weight'],
meta_keys=[
'image_file', 'joints_3d', 'joints_3d_visible', 'center', 'scale',
'rotation', 'bbox_score', 'flip_pairs'
]),
]
val_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(
type='Collect',
keys=['img'],
meta_keys=[
'image_file', 'center', 'scale', 'rotation', 'bbox_score',
'flip_pairs'
]),
]
test_pipeline = val_pipeline
data_root = 'data/coco'
data = dict(
samples_per_gpu=32,
workers_per_gpu=2,
val_dataloader=dict(samples_per_gpu=32),
test_dataloader=dict(samples_per_gpu=32),
train=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_train2017.json',
img_prefix=f'{data_root}/train2017/',
data_cfg=data_cfg,
pipeline=train_pipeline,
dataset_info={{_base_.dataset_info}}),
val=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=val_pipeline,
dataset_info={{_base_.dataset_info}}),
test=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=test_pipeline,
dataset_info={{_base_.dataset_info}}),
)
configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnext152_coco_256x192.py 0 → 100755
View file @ 37c8cebc
_base_ = [
'../../../../_base_/default_runtime.py',
'../../../../_base_/datasets/coco.py'
]
evaluation = dict(interval=10, metric='mAP', save_best='AP')
optimizer = dict(
type='Adam',
lr=5e-4,
)
optimizer_config = dict(grad_clip=None)
# learning policy
lr_config = dict(
policy='step',
warmup='linear',
warmup_iters=500,
warmup_ratio=0.001,
step=[170, 200])
total_epochs = 210
channel_cfg = dict(
num_output_channels=17,
dataset_joints=17,
dataset_channel=[
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16],
],
inference_channel=[
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16
])
# model settings
model = dict(
type='TopDown',
pretrained='mmcls://resnext152_32x4d',
backbone=dict(type='ResNeXt', depth=152),
keypoint_head=dict(
type='TopdownHeatmapSimpleHead',
in_channels=2048,
out_channels=channel_cfg['num_output_channels'],
loss_keypoint=dict(type='JointsMSELoss', use_target_weight=True)),
train_cfg=dict(),
test_cfg=dict(
flip_test=True,
post_process='default',
shift_heatmap=True,
modulate_kernel=11))
data_cfg = dict(
image_size=[192, 256],
heatmap_size=[48, 64],
num_output_channels=channel_cfg['num_output_channels'],
num_joints=channel_cfg['dataset_joints'],
dataset_channel=channel_cfg['dataset_channel'],
inference_channel=channel_cfg['inference_channel'],
soft_nms=False,
nms_thr=1.0,
oks_thr=0.9,
vis_thr=0.2,
use_gt_bbox=False,
det_bbox_thr=0.0,
bbox_file='data/coco/person_detection_results/'
'COCO_val2017_detections_AP_H_56_person.json',
)
train_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownRandomShiftBboxCenter', shift_factor=0.16, prob=0.3),
dict(type='TopDownRandomFlip', flip_prob=0.5),
dict(
type='TopDownHalfBodyTransform',
num_joints_half_body=8,
prob_half_body=0.3),
dict(
type='TopDownGetRandomScaleRotation', rot_factor=40, scale_factor=0.5),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(type='TopDownGenerateTarget', sigma=2),
dict(
type='Collect',
keys=['img', 'target', 'target_weight'],
meta_keys=[
'image_file', 'joints_3d', 'joints_3d_visible', 'center', 'scale',
'rotation', 'bbox_score', 'flip_pairs'
]),
]
val_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(
type='Collect',
keys=['img'],
meta_keys=[
'image_file', 'center', 'scale', 'rotation', 'bbox_score',
'flip_pairs'
]),
]
test_pipeline = val_pipeline
data_root = 'data/coco'
data = dict(
samples_per_gpu=32,
workers_per_gpu=2,
val_dataloader=dict(samples_per_gpu=32),
test_dataloader=dict(samples_per_gpu=32),
train=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_train2017.json',
img_prefix=f'{data_root}/train2017/',
data_cfg=data_cfg,
pipeline=train_pipeline,
dataset_info={{_base_.dataset_info}}),
val=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=val_pipeline,
dataset_info={{_base_.dataset_info}}),
test=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=test_pipeline,
dataset_info={{_base_.dataset_info}}),
)
configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnext152_coco_384x288.py 0 → 100755
View file @ 37c8cebc
_base_ = [
'../../../../_base_/default_runtime.py',
'../../../../_base_/datasets/coco.py'
]
evaluation = dict(interval=10, metric='mAP', save_best='AP')
optimizer = dict(
type='Adam',
lr=5e-4,
)
optimizer_config = dict(grad_clip=None)
# learning policy
lr_config = dict(
policy='step',
warmup='linear',
warmup_iters=500,
warmup_ratio=0.001,
step=[170, 200])
total_epochs = 210
channel_cfg = dict(
num_output_channels=17,
dataset_joints=17,
dataset_channel=[
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16],
],
inference_channel=[
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16
])
# model settings
model = dict(
type='TopDown',
pretrained='mmcls://resnext152_32x4d',
backbone=dict(type='ResNeXt', depth=152),
keypoint_head=dict(
type='TopdownHeatmapSimpleHead',
in_channels=2048,
out_channels=channel_cfg['num_output_channels'],
loss_keypoint=dict(type='JointsMSELoss', use_target_weight=True)),
train_cfg=dict(),
test_cfg=dict(
flip_test=True,
post_process='default',
shift_heatmap=True,
modulate_kernel=11))
data_cfg = dict(
image_size=[288, 384],
heatmap_size=[72, 96],
num_output_channels=channel_cfg['num_output_channels'],
num_joints=channel_cfg['dataset_joints'],
dataset_channel=channel_cfg['dataset_channel'],
inference_channel=channel_cfg['inference_channel'],
soft_nms=False,
nms_thr=1.0,
oks_thr=0.9,
vis_thr=0.2,
use_gt_bbox=False,
det_bbox_thr=0.0,
bbox_file='data/coco/person_detection_results/'
'COCO_val2017_detections_AP_H_56_person.json',
)
train_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownRandomShiftBboxCenter', shift_factor=0.16, prob=0.3),
dict(type='TopDownRandomFlip', flip_prob=0.5),
dict(
type='TopDownHalfBodyTransform',
num_joints_half_body=8,
prob_half_body=0.3),
dict(
type='TopDownGetRandomScaleRotation', rot_factor=40, scale_factor=0.5),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(type='TopDownGenerateTarget', sigma=3),
dict(
type='Collect',
keys=['img', 'target', 'target_weight'],
meta_keys=[
'image_file', 'joints_3d', 'joints_3d_visible', 'center', 'scale',
'rotation', 'bbox_score', 'flip_pairs'
]),
]
val_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(
type='Collect',
keys=['img'],
meta_keys=[
'image_file', 'center', 'scale', 'rotation', 'bbox_score',
'flip_pairs'
]),
]
test_pipeline = val_pipeline
data_root = 'data/coco'
data = dict(
samples_per_gpu=48,
workers_per_gpu=2,
val_dataloader=dict(samples_per_gpu=32),
test_dataloader=dict(samples_per_gpu=32),
train=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_train2017.json',
img_prefix=f'{data_root}/train2017/',
data_cfg=data_cfg,
pipeline=train_pipeline,
dataset_info={{_base_.dataset_info}}),
val=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=val_pipeline,
dataset_info={{_base_.dataset_info}}),
test=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=test_pipeline,
dataset_info={{_base_.dataset_info}}),
)
configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnext50_coco_256x192.py 0 → 100755
View file @ 37c8cebc
_base_ = [
'../../../../_base_/default_runtime.py',
'../../../../_base_/datasets/coco.py'
]
evaluation = dict(interval=10, metric='mAP', save_best='AP')
optimizer = dict(
type='Adam',
lr=5e-4,
)
optimizer_config = dict(grad_clip=None)
# learning policy
lr_config = dict(
policy='step',
warmup='linear',
warmup_iters=500,
warmup_ratio=0.001,
step=[170, 200])
total_epochs = 210
channel_cfg = dict(
num_output_channels=17,
dataset_joints=17,
dataset_channel=[
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16],
],
inference_channel=[
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16
])
# model settings
model = dict(
type='TopDown',
pretrained='mmcls://resnext50_32x4d',
backbone=dict(type='ResNeXt', depth=50),
keypoint_head=dict(
type='TopdownHeatmapSimpleHead',
in_channels=2048,
out_channels=channel_cfg['num_output_channels'],
loss_keypoint=dict(type='JointsMSELoss', use_target_weight=True)),
train_cfg=dict(),
test_cfg=dict(
flip_test=True,
post_process='default',
shift_heatmap=True,
modulate_kernel=11))
data_cfg = dict(
image_size=[192, 256],
heatmap_size=[48, 64],
num_output_channels=channel_cfg['num_output_channels'],
num_joints=channel_cfg['dataset_joints'],
dataset_channel=channel_cfg['dataset_channel'],
inference_channel=channel_cfg['inference_channel'],
soft_nms=False,
nms_thr=1.0,
oks_thr=0.9,
vis_thr=0.2,
use_gt_bbox=False,
det_bbox_thr=0.0,
bbox_file='data/coco/person_detection_results/'
'COCO_val2017_detections_AP_H_56_person.json',
)
train_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownRandomShiftBboxCenter', shift_factor=0.16, prob=0.3),
dict(type='TopDownRandomFlip', flip_prob=0.5),
dict(
type='TopDownHalfBodyTransform',
num_joints_half_body=8,
prob_half_body=0.3),
dict(
type='TopDownGetRandomScaleRotation', rot_factor=40, scale_factor=0.5),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(type='TopDownGenerateTarget', sigma=2),
dict(
type='Collect',
keys=['img', 'target', 'target_weight'],
meta_keys=[
'image_file', 'joints_3d', 'joints_3d_visible', 'center', 'scale',
'rotation', 'bbox_score', 'flip_pairs'
]),
]
val_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(
type='Collect',
keys=['img'],
meta_keys=[
'image_file', 'center', 'scale', 'rotation', 'bbox_score',
'flip_pairs'
]),
]
test_pipeline = val_pipeline
data_root = 'data/coco'
data = dict(
samples_per_gpu=64,
workers_per_gpu=2,
val_dataloader=dict(samples_per_gpu=32),
test_dataloader=dict(samples_per_gpu=32),
train=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_train2017.json',
img_prefix=f'{data_root}/train2017/',
data_cfg=data_cfg,
pipeline=train_pipeline,
dataset_info={{_base_.dataset_info}}),
val=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=val_pipeline,
dataset_info={{_base_.dataset_info}}),
test=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=test_pipeline,
dataset_info={{_base_.dataset_info}}),
)
configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnext50_coco_384x288.py 0 → 100755
View file @ 37c8cebc
_base_ = [
'../../../../_base_/default_runtime.py',
'../../../../_base_/datasets/coco.py'
]
evaluation = dict(interval=10, metric='mAP', save_best='AP')
optimizer = dict(
type='Adam',
lr=5e-4,
)
optimizer_config = dict(grad_clip=None)
# learning policy
lr_config = dict(
policy='step',
warmup='linear',
warmup_iters=500,
warmup_ratio=0.001,
step=[170, 200])
total_epochs = 210
channel_cfg = dict(
num_output_channels=17,
dataset_joints=17,
dataset_channel=[
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16],
],
inference_channel=[
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16
])
# model settings
model = dict(
type='TopDown',
pretrained='mmcls://resnext50_32x4d',
backbone=dict(type='ResNeXt', depth=50),
keypoint_head=dict(
type='TopdownHeatmapSimpleHead',
in_channels=2048,
out_channels=channel_cfg['num_output_channels'],
loss_keypoint=dict(type='JointsMSELoss', use_target_weight=True)),
train_cfg=dict(),
test_cfg=dict(
flip_test=True,
post_process='default',
shift_heatmap=True,
modulate_kernel=11))
data_cfg = dict(
image_size=[288, 384],
heatmap_size=[72, 96],
num_output_channels=channel_cfg['num_output_channels'],
num_joints=channel_cfg['dataset_joints'],
dataset_channel=channel_cfg['dataset_channel'],
inference_channel=channel_cfg['inference_channel'],
soft_nms=False,
nms_thr=1.0,
oks_thr=0.9,
vis_thr=0.2,
use_gt_bbox=False,
det_bbox_thr=0.0,
bbox_file='data/coco/person_detection_results/'
'COCO_val2017_detections_AP_H_56_person.json',
)
train_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownRandomShiftBboxCenter', shift_factor=0.16, prob=0.3),
dict(type='TopDownRandomFlip', flip_prob=0.5),
dict(
type='TopDownHalfBodyTransform',
num_joints_half_body=8,
prob_half_body=0.3),
dict(
type='TopDownGetRandomScaleRotation', rot_factor=40, scale_factor=0.5),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(type='TopDownGenerateTarget', sigma=3),
dict(
type='Collect',
keys=['img', 'target', 'target_weight'],
meta_keys=[
'image_file', 'joints_3d', 'joints_3d_visible', 'center', 'scale',
'rotation', 'bbox_score', 'flip_pairs'
]),
]
val_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(
type='Collect',
keys=['img'],
meta_keys=[
'image_file', 'center', 'scale', 'rotation', 'bbox_score',
'flip_pairs'
]),
]
test_pipeline = val_pipeline
data_root = 'data/coco'
data = dict(
samples_per_gpu=64,
workers_per_gpu=2,
val_dataloader=dict(samples_per_gpu=32),
test_dataloader=dict(samples_per_gpu=32),
train=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_train2017.json',
img_prefix=f'{data_root}/train2017/',
data_cfg=data_cfg,
pipeline=train_pipeline,
dataset_info={{_base_.dataset_info}}),
val=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=val_pipeline,
dataset_info={{_base_.dataset_info}}),
test=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=test_pipeline,
dataset_info={{_base_.dataset_info}}),
)
configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnext_coco.md 0 → 100755
View file @ 37c8cebc
<!-- [BACKBONE] -->
<details>
<summary align="right"><a href="http://openaccess.thecvf.com/content_cvpr_2017/html/Xie_Aggregated_Residual_Transformations_CVPR_2017_paper.html">ResNext (CVPR'2017)</a></summary>
```bibtex
@inproceedings{xie2017aggregated,
title={Aggregated residual transformations for deep neural networks},
author={Xie, Saining and Girshick, Ross and Doll{\'a}r, Piotr and Tu, Zhuowen and He, Kaiming},
booktitle={Proceedings of the IEEE conference on computer vision and pattern recognition},
pages={1492--1500},
year={2017}
}
```
</details>
<!-- [DATASET] -->
<details>
<summary align="right"><a href="https://link.springer.com/chapter/10.1007/978-3-319-10602-1_48">COCO (ECCV'2014)</a></summary>
```bibtex
@inproceedings{lin2014microsoft,
title={Microsoft coco: Common objects in context},
author={Lin, Tsung-Yi and Maire, Michael and Belongie, Serge and Hays, James and Perona, Pietro and Ramanan, Deva and Doll{\'a}r, Piotr and Zitnick, C Lawrence},
booktitle={European conference on computer vision},
pages={740--755},
year={2014},
organization={Springer}
}
```
</details>
Results on COCO val2017 with detector having human AP of 56.4 on COCO val2017 dataset
| Arch | Input Size | AP | AP<sup>50</sup> | AP<sup>75</sup> | AR | AR<sup>50</sup> | ckpt | log |
| :-------------------------------------------- | :--------: | :---: | :-------------: | :-------------: | :---: | :-------------: | :-------------------------------------------: | :-------------------------------------------: |
| [pose_resnext_50](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnext50_coco_256x192.py) | 256x192 | 0.714 | 0.898 | 0.789 | 0.771 | 0.937 | [ckpt](https://download.openmmlab.com/mmpose/top_down/resnext/resnext50_coco_256x192-dcff15f6_20200727.pth) | [log](https://download.openmmlab.com/mmpose/top_down/resnext/resnext50_coco_256x192_20200727.log.json) |
| [pose_resnext_50](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnext50_coco_384x288.py) | 384x288 | 0.724 | 0.899 | 0.794 | 0.777 | 0.935 | [ckpt](https://download.openmmlab.com/mmpose/top_down/resnext/resnext50_coco_384x288-412c848f_20200727.pth) | [log](https://download.openmmlab.com/mmpose/top_down/resnext/resnext50_coco_384x288_20200727.log.json) |
| [pose_resnext_101](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnext101_coco_256x192.py) | 256x192 | 0.726 | 0.900 | 0.801 | 0.782 | 0.940 | [ckpt](https://download.openmmlab.com/mmpose/top_down/resnext/resnext101_coco_256x192-c7eba365_20200727.pth) | [log](https://download.openmmlab.com/mmpose/top_down/resnext/resnext101_coco_256x192_20200727.log.json) |
| [pose_resnext_101](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnext101_coco_384x288.py) | 384x288 | 0.743 | 0.903 | 0.815 | 0.795 | 0.939 | [ckpt](https://download.openmmlab.com/mmpose/top_down/resnext/resnext101_coco_384x288-f5eabcd6_20200727.pth) | [log](https://download.openmmlab.com/mmpose/top_down/resnext/resnext101_coco_384x288_20200727.log.json) |
| [pose_resnext_152](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnext152_coco_256x192.py) | 256x192 | 0.730 | 0.904 | 0.808 | 0.786 | 0.940 | [ckpt](https://download.openmmlab.com/mmpose/top_down/resnext/resnext152_coco_256x192-102449aa_20200727.pth) | [log](https://download.openmmlab.com/mmpose/top_down/resnext/resnext152_coco_256x192_20200727.log.json) |
| [pose_resnext_152](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnext152_coco_384x288.py) | 384x288 | 0.742 | 0.902 | 0.810 | 0.794 | 0.939 | [ckpt](https://download.openmmlab.com/mmpose/top_down/resnext/resnext152_coco_384x288-806176df_20200727.pth) | [log](https://download.openmmlab.com/mmpose/top_down/resnext/resnext152_coco_384x288_20200727.log.json) |
configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnext_coco.yml 0 → 100755
View file @ 37c8cebc
Collections:
- Name: ResNext
Paper:
Title: Aggregated residual transformations for deep neural networks
URL: http://openaccess.thecvf.com/content_cvpr_2017/html/Xie_Aggregated_Residual_Transformations_CVPR_2017_paper.html
README: https://github.com/open-mmlab/mmpose/blob/master/docs/en/papers/backbones/resnext.md
Models:
- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnext50_coco_256x192.py
In Collection: ResNext
Metadata:
Architecture: &id001
- ResNext
Training Data: COCO
Name: topdown_heatmap_resnext50_coco_256x192
Results:
- Dataset: COCO
Metrics:
AP: 0.714
AP@0.5: 0.898
AP@0.75: 0.789
AR: 0.771
AR@0.5: 0.937
Task: Body 2D Keypoint
Weights: https://download.openmmlab.com/mmpose/top_down/resnext/resnext50_coco_256x192-dcff15f6_20200727.pth
- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnext50_coco_384x288.py
In Collection: ResNext
Metadata:
Architecture: *id001
Training Data: COCO
Name: topdown_heatmap_resnext50_coco_384x288
Results:
- Dataset: COCO
Metrics:
AP: 0.724
AP@0.5: 0.899
AP@0.75: 0.794
AR: 0.777
AR@0.5: 0.935
Task: Body 2D Keypoint
Weights: https://download.openmmlab.com/mmpose/top_down/resnext/resnext50_coco_384x288-412c848f_20200727.pth
- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnext101_coco_256x192.py
In Collection: ResNext
Metadata:
Architecture: *id001
Training Data: COCO
Name: topdown_heatmap_resnext101_coco_256x192
Results:
- Dataset: COCO
Metrics:
AP: 0.726
AP@0.5: 0.9
AP@0.75: 0.801
AR: 0.782
AR@0.5: 0.94
Task: Body 2D Keypoint
Weights: https://download.openmmlab.com/mmpose/top_down/resnext/resnext101_coco_256x192-c7eba365_20200727.pth
- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnext101_coco_384x288.py
In Collection: ResNext
Metadata:
Architecture: *id001
Training Data: COCO
Name: topdown_heatmap_resnext101_coco_384x288
Results:
- Dataset: COCO
Metrics:
AP: 0.743
AP@0.5: 0.903
AP@0.75: 0.815
AR: 0.795
AR@0.5: 0.939
Task: Body 2D Keypoint
Weights: https://download.openmmlab.com/mmpose/top_down/resnext/resnext101_coco_384x288-f5eabcd6_20200727.pth
- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnext152_coco_256x192.py
In Collection: ResNext
Metadata:
Architecture: *id001
Training Data: COCO
Name: topdown_heatmap_resnext152_coco_256x192
Results:
- Dataset: COCO
Metrics:
AP: 0.73
AP@0.5: 0.904
AP@0.75: 0.808
AR: 0.786
AR@0.5: 0.94
Task: Body 2D Keypoint
Weights: https://download.openmmlab.com/mmpose/top_down/resnext/resnext152_coco_256x192-102449aa_20200727.pth
- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/resnext152_coco_384x288.py
In Collection: ResNext
Metadata:
Architecture: *id001
Training Data: COCO
Name: topdown_heatmap_resnext152_coco_384x288
Results:
- Dataset: COCO
Metrics:
AP: 0.742
AP@0.5: 0.902
AP@0.75: 0.81
AR: 0.794
AR@0.5: 0.939
Task: Body 2D Keypoint
Weights: https://download.openmmlab.com/mmpose/top_down/resnext/resnext152_coco_384x288-806176df_20200727.pth
configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/rsn18_coco_256x192.py 0 → 100755
View file @ 37c8cebc
_base_ = [
'../../../../_base_/default_runtime.py',
'../../../../_base_/datasets/coco.py'
]
evaluation = dict(interval=10, metric='mAP', save_best='AP')
optimizer = dict(
type='Adam',
lr=2e-2,
)
optimizer_config = dict(grad_clip=None)
# learning policy
lr_config = dict(
policy='step',
warmup='linear',
warmup_iters=500,
warmup_ratio=0.001,
step=[170, 190, 200])
total_epochs = 210
log_config = dict(
interval=50, hooks=[
dict(type='TextLoggerHook'),
])
channel_cfg = dict(
num_output_channels=17,
dataset_joints=17,
dataset_channel=[
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16],
],
inference_channel=[
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16
])
# model settings
model = dict(
type='TopDown',
pretrained=None,
backbone=dict(
type='RSN',
unit_channels=256,
num_stages=1,
num_units=4,
num_blocks=[2, 2, 2, 2],
num_steps=4,
norm_cfg=dict(type='BN')),
keypoint_head=dict(
type='TopdownHeatmapMSMUHead',
out_shape=(64, 48),
unit_channels=256,
out_channels=channel_cfg['num_output_channels'],
num_stages=1,
num_units=4,
use_prm=False,
norm_cfg=dict(type='BN'),
loss_keypoint=[
dict(
type='JointsMSELoss', use_target_weight=True, loss_weight=0.25)
] * 3 + [
dict(
type='JointsOHKMMSELoss',
use_target_weight=True,
loss_weight=1.)
]),
train_cfg=dict(),
test_cfg=dict(
flip_test=True,
post_process='megvii',
shift_heatmap=False,
modulate_kernel=5))
data_cfg = dict(
image_size=[192, 256],
heatmap_size=[48, 64],
num_output_channels=channel_cfg['num_output_channels'],
num_joints=channel_cfg['dataset_joints'],
dataset_channel=channel_cfg['dataset_channel'],
inference_channel=channel_cfg['inference_channel'],
soft_nms=False,
use_nms=False,
nms_thr=1.0,
oks_thr=0.9,
vis_thr=0.2,
use_gt_bbox=False,
det_bbox_thr=0.0,
bbox_file='data/coco/person_detection_results/'
'COCO_val2017_detections_AP_H_56_person.json',
)
train_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownRandomShiftBboxCenter', shift_factor=0.16, prob=0.3),
dict(type='TopDownRandomFlip', flip_prob=0.5),
dict(
type='TopDownHalfBodyTransform',
num_joints_half_body=8,
prob_half_body=0.3),
dict(
type='TopDownGetRandomScaleRotation', rot_factor=40, scale_factor=0.5),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(
type='TopDownGenerateTarget',
kernel=[(11, 11), (9, 9), (7, 7), (5, 5)],
encoding='Megvii'),
dict(
type='Collect',
keys=['img', 'target', 'target_weight'],
meta_keys=[
'image_file', 'joints_3d', 'joints_3d_visible', 'center', 'scale',
'rotation', 'bbox_score', 'flip_pairs'
]),
]
val_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(
type='Collect',
keys=[
'img',
],
meta_keys=[
'image_file', 'center', 'scale', 'rotation', 'bbox_score',
'flip_pairs'
]),
]
test_pipeline = val_pipeline
data_root = 'data/coco'
data = dict(
samples_per_gpu=32,
workers_per_gpu=4,
train=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_train2017.json',
img_prefix=f'{data_root}/train2017/',
data_cfg=data_cfg,
pipeline=train_pipeline,
dataset_info={{_base_.dataset_info}}),
val=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=val_pipeline,
dataset_info={{_base_.dataset_info}}),
test=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=test_pipeline,
dataset_info={{_base_.dataset_info}}),
)
configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/rsn50_coco_256x192.py 0 → 100755
View file @ 37c8cebc
_base_ = [
'../../../../_base_/default_runtime.py',
'../../../../_base_/datasets/coco.py'
]
evaluation = dict(interval=10, metric='mAP', save_best='AP')
optimizer = dict(
type='Adam',
lr=5e-3,
)
optimizer_config = dict(grad_clip=None)
# learning policy
lr_config = dict(
policy='step',
warmup='linear',
warmup_iters=500,
warmup_ratio=0.001,
step=[170, 200])
total_epochs = 210
log_config = dict(
interval=50, hooks=[
dict(type='TextLoggerHook'),
])
channel_cfg = dict(
num_output_channels=17,
dataset_joints=17,
dataset_channel=[
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16],
],
inference_channel=[
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16
])
# model settings
model = dict(
type='TopDown',
pretrained=None,
backbone=dict(
type='RSN',
unit_channels=256,
num_stages=1,
num_units=4,
num_blocks=[3, 4, 6, 3],
num_steps=4,
norm_cfg=dict(type='BN')),
keypoint_head=dict(
type='TopdownHeatmapMSMUHead',
out_shape=(64, 48),
unit_channels=256,
out_channels=channel_cfg['num_output_channels'],
num_stages=1,
num_units=4,
use_prm=False,
norm_cfg=dict(type='BN'),
loss_keypoint=[
dict(
type='JointsMSELoss', use_target_weight=True, loss_weight=0.25)
] * 3 + [
dict(
type='JointsOHKMMSELoss',
use_target_weight=True,
loss_weight=1.)
]),
train_cfg=dict(),
test_cfg=dict(
flip_test=True,
post_process='megvii',
shift_heatmap=False,
modulate_kernel=5))
data_cfg = dict(
image_size=[192, 256],
heatmap_size=[48, 64],
num_output_channels=channel_cfg['num_output_channels'],
num_joints=channel_cfg['dataset_joints'],
dataset_channel=channel_cfg['dataset_channel'],
inference_channel=channel_cfg['inference_channel'],
soft_nms=False,
use_nms=False,
nms_thr=1.0,
oks_thr=0.9,
vis_thr=0.2,
use_gt_bbox=False,
det_bbox_thr=0.0,
bbox_file='data/coco/person_detection_results/'
'COCO_val2017_detections_AP_H_56_person.json',
)
train_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownRandomShiftBboxCenter', shift_factor=0.16, prob=0.3),
dict(type='TopDownRandomFlip', flip_prob=0.5),
dict(
type='TopDownHalfBodyTransform',
num_joints_half_body=8,
prob_half_body=0.3),
dict(
type='TopDownGetRandomScaleRotation', rot_factor=40, scale_factor=0.5),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(
type='TopDownGenerateTarget',
kernel=[(11, 11), (9, 9), (7, 7), (5, 5)],
encoding='Megvii'),
dict(
type='Collect',
keys=['img', 'target', 'target_weight'],
meta_keys=[
'image_file', 'joints_3d', 'joints_3d_visible', 'center', 'scale',
'rotation', 'bbox_score', 'flip_pairs'
]),
]
val_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(
type='Collect',
keys=[
'img',
],
meta_keys=[
'image_file', 'center', 'scale', 'rotation', 'bbox_score',
'flip_pairs'
]),
]
test_pipeline = val_pipeline
data_root = 'data/coco'
data = dict(
samples_per_gpu=32,
workers_per_gpu=4,
train=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_train2017.json',
img_prefix=f'{data_root}/train2017/',
data_cfg=data_cfg,
pipeline=train_pipeline,
dataset_info={{_base_.dataset_info}}),
val=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=val_pipeline,
dataset_info={{_base_.dataset_info}}),
test=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=test_pipeline,
dataset_info={{_base_.dataset_info}}),
)
configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/rsn_coco.md 0 → 100755
View file @ 37c8cebc
<!-- [ALGORITHM] -->
<details>
<summary align="right"><a href="https://link.springer.com/chapter/10.1007/978-3-030-58580-8_27">RSN (ECCV'2020)</a></summary>
```bibtex
@misc{cai2020learning,
title={Learning Delicate Local Representations for Multi-Person Pose Estimation},
author={Yuanhao Cai and Zhicheng Wang and Zhengxiong Luo and Binyi Yin and Angang Du and Haoqian Wang and Xinyu Zhou and Erjin Zhou and Xiangyu Zhang and Jian Sun},
year={2020},
eprint={2003.04030},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
</details>
<!-- [DATASET] -->
<details>
<summary align="right"><a href="https://link.springer.com/chapter/10.1007/978-3-319-10602-1_48">COCO (ECCV'2014)</a></summary>
```bibtex
@inproceedings{lin2014microsoft,
title={Microsoft coco: Common objects in context},
author={Lin, Tsung-Yi and Maire, Michael and Belongie, Serge and Hays, James and Perona, Pietro and Ramanan, Deva and Doll{\'a}r, Piotr and Zitnick, C Lawrence},
booktitle={European conference on computer vision},
pages={740--755},
year={2014},
organization={Springer}
}
```
</details>
Results on COCO val2017 with detector having human AP of 56.4 on COCO val2017 dataset
| Arch | Input Size | AP | AP<sup>50</sup> | AP<sup>75</sup> | AR | AR<sup>50</sup> | ckpt | log |
| :-------------------------------------------- | :--------: | :---: | :-------------: | :-------------: | :---: | :-------------: | :-------------------------------------------: | :-------------------------------------------: |
| [rsn_18](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/rsn18_coco_256x192.py) | 256x192 | 0.704 | 0.887 | 0.779 | 0.771 | 0.926 | [ckpt](https://download.openmmlab.com/mmpose/top_down/rsn/rsn18_coco_256x192-72f4b4a7_20201127.pth) | [log](https://download.openmmlab.com/mmpose/top_down/rsn/rsn18_coco_256x192_20201127.log.json) |
| [rsn_50](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/rsn50_coco_256x192.py) | 256x192 | 0.723 | 0.896 | 0.800 | 0.788 | 0.934 | [ckpt](https://download.openmmlab.com/mmpose/top_down/rsn/rsn50_coco_256x192-72ffe709_20201127.pth) | [log](https://download.openmmlab.com/mmpose/top_down/rsn/rsn50_coco_256x192_20201127.log.json) |
| [2xrsn_50](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/2xrsn50_coco_256x192.py) | 256x192 | 0.745 | 0.899 | 0.818 | 0.809 | 0.939 | [ckpt](https://download.openmmlab.com/mmpose/top_down/rsn/2xrsn50_coco_256x192-50648f0e_20201127.pth) | [log](https://download.openmmlab.com/mmpose/top_down/rsn/2xrsn50_coco_256x192_20201127.log.json) |
| [3xrsn_50](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/3xrsn50_coco_256x192.py) | 256x192 | 0.750 | 0.900 | 0.823 | 0.813 | 0.940 | [ckpt](https://download.openmmlab.com/mmpose/top_down/rsn/3xrsn50_coco_256x192-58f57a68_20201127.pth) | [log](https://download.openmmlab.com/mmpose/top_down/rsn/3xrsn50_coco_256x192_20201127.log.json) |
configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/rsn_coco.yml 0 → 100755
View file @ 37c8cebc
Collections:
- Name: RSN
Paper:
Title: Learning Delicate Local Representations for Multi-Person Pose Estimation
URL: https://link.springer.com/chapter/10.1007/978-3-030-58580-8_27
README: https://github.com/open-mmlab/mmpose/blob/master/docs/en/papers/backbones/rsn.md
Models:
- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/rsn18_coco_256x192.py
In Collection: RSN
Metadata:
Architecture: &id001
- RSN
Training Data: COCO
Name: topdown_heatmap_rsn18_coco_256x192
Results:
- Dataset: COCO
Metrics:
AP: 0.704
AP@0.5: 0.887
AP@0.75: 0.779
AR: 0.771
AR@0.5: 0.926
Task: Body 2D Keypoint
Weights: https://download.openmmlab.com/mmpose/top_down/rsn/rsn18_coco_256x192-72f4b4a7_20201127.pth
- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/rsn50_coco_256x192.py
In Collection: RSN
Metadata:
Architecture: *id001
Training Data: COCO
Name: topdown_heatmap_rsn50_coco_256x192
Results:
- Dataset: COCO
Metrics:
AP: 0.723
AP@0.5: 0.896
AP@0.75: 0.8
AR: 0.788
AR@0.5: 0.934
Task: Body 2D Keypoint
Weights: https://download.openmmlab.com/mmpose/top_down/rsn/rsn50_coco_256x192-72ffe709_20201127.pth
- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/2xrsn50_coco_256x192.py
In Collection: RSN
Metadata:
Architecture: *id001
Training Data: COCO
Name: topdown_heatmap_2xrsn50_coco_256x192
Results:
- Dataset: COCO
Metrics:
AP: 0.745
AP@0.5: 0.899
AP@0.75: 0.818
AR: 0.809
AR@0.5: 0.939
Task: Body 2D Keypoint
Weights: https://download.openmmlab.com/mmpose/top_down/rsn/2xrsn50_coco_256x192-50648f0e_20201127.pth
- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/3xrsn50_coco_256x192.py
In Collection: RSN
Metadata:
Architecture: *id001
Training Data: COCO
Name: topdown_heatmap_3xrsn50_coco_256x192
Results:
- Dataset: COCO
Metrics:
AP: 0.75
AP@0.5: 0.9
AP@0.75: 0.823
AR: 0.813
AR@0.5: 0.94
Task: Body 2D Keypoint
Weights: https://download.openmmlab.com/mmpose/top_down/rsn/3xrsn50_coco_256x192-58f57a68_20201127.pth
configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/scnet101_coco_256x192.py 0 → 100755
View file @ 37c8cebc
_base_ = [
'../../../../_base_/default_runtime.py',
'../../../../_base_/datasets/coco.py'
]
evaluation = dict(interval=10, metric='mAP', save_best='AP')
optimizer = dict(
type='Adam',
lr=5e-4,
)
optimizer_config = dict(grad_clip=None)
# learning policy
lr_config = dict(
policy='step',
warmup='linear',
warmup_iters=500,
warmup_ratio=0.001,
step=[170, 200])
total_epochs = 210
channel_cfg = dict(
num_output_channels=17,
dataset_joints=17,
dataset_channel=[
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16],
],
inference_channel=[
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16
])
# model settings
model = dict(
type='TopDown',
pretrained='https://download.openmmlab.com/mmpose/'
'pretrain_models/scnet101-94250a77.pth',
backbone=dict(type='SCNet', depth=101),
keypoint_head=dict(
type='TopdownHeatmapSimpleHead',
in_channels=2048,
out_channels=channel_cfg['num_output_channels'],
loss_keypoint=dict(type='JointsMSELoss', use_target_weight=True)),
train_cfg=dict(),
test_cfg=dict(
flip_test=True,
post_process='default',
shift_heatmap=True,
modulate_kernel=11))
data_cfg = dict(
image_size=[192, 256],
heatmap_size=[48, 64],
num_output_channels=channel_cfg['num_output_channels'],
num_joints=channel_cfg['dataset_joints'],
dataset_channel=channel_cfg['dataset_channel'],
inference_channel=channel_cfg['inference_channel'],
soft_nms=False,
nms_thr=1.0,
oks_thr=0.9,
vis_thr=0.2,
use_gt_bbox=False,
det_bbox_thr=0.0,
bbox_file='data/coco/person_detection_results/'
'COCO_val2017_detections_AP_H_56_person.json',
)
train_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownRandomShiftBboxCenter', shift_factor=0.16, prob=0.3),
dict(type='TopDownRandomFlip', flip_prob=0.5),
dict(
type='TopDownHalfBodyTransform',
num_joints_half_body=8,
prob_half_body=0.3),
dict(
type='TopDownGetRandomScaleRotation', rot_factor=40, scale_factor=0.5),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(type='TopDownGenerateTarget', sigma=2),
dict(
type='Collect',
keys=['img', 'target', 'target_weight'],
meta_keys=[
'image_file', 'joints_3d', 'joints_3d_visible', 'center', 'scale',
'rotation', 'bbox_score', 'flip_pairs'
]),
]
val_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(
type='Collect',
keys=['img'],
meta_keys=[
'image_file', 'center', 'scale', 'rotation', 'bbox_score',
'flip_pairs'
]),
]
test_pipeline = val_pipeline
data_root = 'data/coco'
data = dict(
samples_per_gpu=32,
workers_per_gpu=1,
train=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_train2017.json',
img_prefix=f'{data_root}/train2017/',
data_cfg=data_cfg,
pipeline=train_pipeline,
dataset_info={{_base_.dataset_info}}),
val=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=val_pipeline,
dataset_info={{_base_.dataset_info}}),
test=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=test_pipeline,
dataset_info={{_base_.dataset_info}}),
)
configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/scnet101_coco_384x288.py 0 → 100755
View file @ 37c8cebc
_base_ = [
'../../../../_base_/default_runtime.py',
'../../../../_base_/datasets/coco.py'
]
evaluation = dict(interval=10, metric='mAP', save_best='AP')
optimizer = dict(
type='Adam',
lr=5e-4,
)
optimizer_config = dict(grad_clip=None)
# learning policy
lr_config = dict(
policy='step',
warmup='linear',
warmup_iters=500,
warmup_ratio=0.001,
step=[170, 200])
total_epochs = 210
channel_cfg = dict(
num_output_channels=17,
dataset_joints=17,
dataset_channel=[
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16],
],
inference_channel=[
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16
])
# model settings
model = dict(
type='TopDown',
pretrained='https://download.openmmlab.com/mmpose/'
'pretrain_models/scnet101-94250a77.pth',
backbone=dict(type='SCNet', depth=101),
keypoint_head=dict(
type='TopdownHeatmapSimpleHead',
in_channels=2048,
out_channels=channel_cfg['num_output_channels'],
loss_keypoint=dict(type='JointsMSELoss', use_target_weight=True)),
train_cfg=dict(),
test_cfg=dict(
flip_test=True,
post_process='default',
shift_heatmap=True,
modulate_kernel=11))
data_cfg = dict(
image_size=[288, 384],
heatmap_size=[72, 96],
num_output_channels=channel_cfg['num_output_channels'],
num_joints=channel_cfg['dataset_joints'],
dataset_channel=channel_cfg['dataset_channel'],
inference_channel=channel_cfg['inference_channel'],
soft_nms=False,
nms_thr=1.0,
oks_thr=0.9,
vis_thr=0.2,
use_gt_bbox=False,
det_bbox_thr=0.0,
bbox_file='data/coco/person_detection_results/'
'COCO_val2017_detections_AP_H_56_person.json',
)
train_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownRandomShiftBboxCenter', shift_factor=0.16, prob=0.3),
dict(type='TopDownRandomFlip', flip_prob=0.5),
dict(
type='TopDownHalfBodyTransform',
num_joints_half_body=8,
prob_half_body=0.3),
dict(
type='TopDownGetRandomScaleRotation', rot_factor=40, scale_factor=0.5),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(type='TopDownGenerateTarget', sigma=3),
dict(
type='Collect',
keys=['img', 'target', 'target_weight'],
meta_keys=[
'image_file', 'joints_3d', 'joints_3d_visible', 'center', 'scale',
'rotation', 'bbox_score', 'flip_pairs'
]),
]
val_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(
type='Collect',
keys=['img'],
meta_keys=[
'image_file', 'center', 'scale', 'rotation', 'bbox_score',
'flip_pairs'
]),
]
test_pipeline = val_pipeline
data_root = 'data/coco'
data = dict(
samples_per_gpu=48,
workers_per_gpu=2,
val_dataloader=dict(samples_per_gpu=32),
test_dataloader=dict(samples_per_gpu=32),
train=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_train2017.json',
img_prefix=f'{data_root}/train2017/',
data_cfg=data_cfg,
pipeline=train_pipeline,
dataset_info={{_base_.dataset_info}}),
val=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=val_pipeline,
dataset_info={{_base_.dataset_info}}),
test=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=test_pipeline,
dataset_info={{_base_.dataset_info}}),
)
configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/scnet50_coco_256x192.py 0 → 100755
View file @ 37c8cebc
_base_ = [
'../../../../_base_/default_runtime.py',
'../../../../_base_/datasets/coco.py'
]
evaluation = dict(interval=10, metric='mAP', save_best='AP')
optimizer = dict(
type='Adam',
lr=5e-4,
)
optimizer_config = dict(grad_clip=None)
# learning policy
lr_config = dict(
policy='step',
warmup='linear',
warmup_iters=500,
warmup_ratio=0.001,
step=[170, 200])
total_epochs = 210
channel_cfg = dict(
num_output_channels=17,
dataset_joints=17,
dataset_channel=[
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16],
],
inference_channel=[
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16
])
# model settings
model = dict(
type='TopDown',
pretrained='https://download.openmmlab.com/mmpose/'
'pretrain_models/scnet50-7ef0a199.pth',
backbone=dict(type='SCNet', depth=50),
keypoint_head=dict(
type='TopdownHeatmapSimpleHead',
in_channels=2048,
out_channels=channel_cfg['num_output_channels'],
loss_keypoint=dict(type='JointsMSELoss', use_target_weight=True)),
train_cfg=dict(),
test_cfg=dict(
flip_test=True,
post_process='default',
shift_heatmap=True,
modulate_kernel=11))
data_cfg = dict(
image_size=[192, 256],
heatmap_size=[48, 64],
num_output_channels=channel_cfg['num_output_channels'],
num_joints=channel_cfg['dataset_joints'],
dataset_channel=channel_cfg['dataset_channel'],
inference_channel=channel_cfg['inference_channel'],
soft_nms=False,
nms_thr=1.0,
oks_thr=0.9,
vis_thr=0.2,
use_gt_bbox=False,
det_bbox_thr=0.0,
bbox_file='data/coco/person_detection_results/'
'COCO_val2017_detections_AP_H_56_person.json',
)
train_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownRandomShiftBboxCenter', shift_factor=0.16, prob=0.3),
dict(type='TopDownRandomFlip', flip_prob=0.5),
dict(
type='TopDownHalfBodyTransform',
num_joints_half_body=8,
prob_half_body=0.3),
dict(
type='TopDownGetRandomScaleRotation', rot_factor=40, scale_factor=0.5),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(type='TopDownGenerateTarget', sigma=2),
dict(
type='Collect',
keys=['img', 'target', 'target_weight'],
meta_keys=[
'image_file', 'joints_3d', 'joints_3d_visible', 'center', 'scale',
'rotation', 'bbox_score', 'flip_pairs'
]),
]
val_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(
type='Collect',
keys=['img'],
meta_keys=[
'image_file', 'center', 'scale', 'rotation', 'bbox_score',
'flip_pairs'
]),
]
test_pipeline = val_pipeline
data_root = 'data/coco'
data = dict(
samples_per_gpu=64,
workers_per_gpu=2,
val_dataloader=dict(samples_per_gpu=32),
test_dataloader=dict(samples_per_gpu=32),
train=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_train2017.json',
img_prefix=f'{data_root}/train2017/',
data_cfg=data_cfg,
pipeline=train_pipeline,
dataset_info={{_base_.dataset_info}}),
val=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=val_pipeline,
dataset_info={{_base_.dataset_info}}),
test=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=test_pipeline,
dataset_info={{_base_.dataset_info}}),
)
configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/scnet50_coco_384x288.py 0 → 100755
View file @ 37c8cebc
_base_ = [
'../../../../_base_/default_runtime.py',
'../../../../_base_/datasets/coco.py'
]
evaluation = dict(interval=10, metric='mAP', save_best='AP')
optimizer = dict(
type='Adam',
lr=5e-4,
)
optimizer_config = dict(grad_clip=None)
# learning policy
lr_config = dict(
policy='step',
warmup='linear',
warmup_iters=500,
warmup_ratio=0.001,
step=[170, 200])
total_epochs = 210
channel_cfg = dict(
num_output_channels=17,
dataset_joints=17,
dataset_channel=[
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16],
],
inference_channel=[
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16
])
# model settings
model = dict(
type='TopDown',
pretrained='https://download.openmmlab.com/mmpose/'
'pretrain_models/scnet50-7ef0a199.pth',
backbone=dict(type='SCNet', depth=50),
keypoint_head=dict(
type='TopdownHeatmapSimpleHead',
in_channels=2048,
out_channels=channel_cfg['num_output_channels'],
loss_keypoint=dict(type='JointsMSELoss', use_target_weight=True)),
train_cfg=dict(),
test_cfg=dict(
flip_test=True,
post_process='default',
shift_heatmap=True,
modulate_kernel=11))
data_cfg = dict(
image_size=[288, 384],
heatmap_size=[72, 96],
num_output_channels=channel_cfg['num_output_channels'],
num_joints=channel_cfg['dataset_joints'],
dataset_channel=channel_cfg['dataset_channel'],
inference_channel=channel_cfg['inference_channel'],
soft_nms=False,
nms_thr=1.0,
oks_thr=0.9,
vis_thr=0.2,
use_gt_bbox=False,
det_bbox_thr=0.0,
bbox_file='data/coco/person_detection_results/'
'COCO_val2017_detections_AP_H_56_person.json',
)
train_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownRandomShiftBboxCenter', shift_factor=0.16, prob=0.3),
dict(type='TopDownRandomFlip', flip_prob=0.5),
dict(
type='TopDownHalfBodyTransform',
num_joints_half_body=8,
prob_half_body=0.3),
dict(
type='TopDownGetRandomScaleRotation', rot_factor=40, scale_factor=0.5),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(type='TopDownGenerateTarget', sigma=3),
dict(
type='Collect',
keys=['img', 'target', 'target_weight'],
meta_keys=[
'image_file', 'joints_3d', 'joints_3d_visible', 'center', 'scale',
'rotation', 'bbox_score', 'flip_pairs'
]),
]
val_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='TopDownGetBboxCenterScale', padding=1.25),
dict(type='TopDownAffine'),
dict(type='ToTensor'),
dict(
type='NormalizeTensor',
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
dict(
type='Collect',
keys=['img'],
meta_keys=[
'image_file', 'center', 'scale', 'rotation', 'bbox_score',
'flip_pairs'
]),
]
test_pipeline = val_pipeline
data_root = 'data/coco'
data = dict(
samples_per_gpu=32,
workers_per_gpu=1,
train=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_train2017.json',
img_prefix=f'{data_root}/train2017/',
data_cfg=data_cfg,
pipeline=train_pipeline,
dataset_info={{_base_.dataset_info}}),
val=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=val_pipeline,
dataset_info={{_base_.dataset_info}}),
test=dict(
type='TopDownCocoDataset',
ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
img_prefix=f'{data_root}/val2017/',
data_cfg=data_cfg,
pipeline=test_pipeline,
dataset_info={{_base_.dataset_info}}),
)
configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/scnet_coco.md 0 → 100755
View file @ 37c8cebc
<!-- [ALGORITHM] -->
<details>
<summary align="right"><a href="http://openaccess.thecvf.com/content_CVPR_2020/html/Liu_Improving_Convolutional_Networks_With_Self-Calibrated_Convolutions_CVPR_2020_paper.html">SCNet (CVPR'2020)</a></summary>
```bibtex
@inproceedings{liu2020improving,
title={Improving Convolutional Networks with Self-Calibrated Convolutions},
author={Liu, Jiang-Jiang and Hou, Qibin and Cheng, Ming-Ming and Wang, Changhu and Feng, Jiashi},
booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
pages={10096--10105},
year={2020}
}
```
</details>
<!-- [DATASET] -->
<details>
<summary align="right"><a href="https://link.springer.com/chapter/10.1007/978-3-319-10602-1_48">COCO (ECCV'2014)</a></summary>
```bibtex
@inproceedings{lin2014microsoft,
title={Microsoft coco: Common objects in context},
author={Lin, Tsung-Yi and Maire, Michael and Belongie, Serge and Hays, James and Perona, Pietro and Ramanan, Deva and Doll{\'a}r, Piotr and Zitnick, C Lawrence},
booktitle={European conference on computer vision},
pages={740--755},
year={2014},
organization={Springer}
}
```
</details>
Results on COCO val2017 with detector having human AP of 56.4 on COCO val2017 dataset
| Arch | Input Size | AP | AP<sup>50</sup> | AP<sup>75</sup> | AR | AR<sup>50</sup> | ckpt | log |
| :-------------------------------------------- | :--------: | :---: | :-------------: | :-------------: | :---: | :-------------: | :-------------------------------------------: | :-------------------------------------------: |
| [pose_scnet_50](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/scnet50_coco_256x192.py) | 256x192 | 0.728 | 0.899 | 0.807 | 0.784 | 0.938 | [ckpt](https://download.openmmlab.com/mmpose/top_down/scnet/scnet50_coco_256x192-6920f829_20200709.pth) | [log](https://download.openmmlab.com/mmpose/top_down/scnet/scnet50_coco_256x192_20200709.log.json) |
| [pose_scnet_50](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/scnet50_coco_384x288.py) | 384x288 | 0.751 | 0.906 | 0.818 | 0.802 | 0.943 | [ckpt](https://download.openmmlab.com/mmpose/top_down/scnet/scnet50_coco_384x288-9cacd0ea_20200709.pth) | [log](https://download.openmmlab.com/mmpose/top_down/scnet/scnet50_coco_384x288_20200709.log.json) |
| [pose_scnet_101](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/scnet101_coco_256x192.py) | 256x192 | 0.733 | 0.903 | 0.813 | 0.790 | 0.941 | [ckpt](https://download.openmmlab.com/mmpose/top_down/scnet/scnet101_coco_256x192-6d348ef9_20200709.pth) | [log](https://download.openmmlab.com/mmpose/top_down/scnet/scnet101_coco_256x192_20200709.log.json) |
| [pose_scnet_101](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/scnet101_coco_384x288.py) | 384x288 | 0.752 | 0.906 | 0.823 | 0.804 | 0.943 | [ckpt](https://download.openmmlab.com/mmpose/top_down/scnet/scnet101_coco_384x288-0b6e631b_20200709.pth) | [log](https://download.openmmlab.com/mmpose/top_down/scnet/scnet101_coco_384x288_20200709.log.json) |
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