add new model

configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/3xmspn50_coco_256x192.py

+_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='torchvision://resnet50',
+    backbone=dict(
+        type='MSPN',
+        unit_channels=256,
+        num_stages=3,
+        num_units=4,
+        num_blocks=[3, 4, 6, 3],
+        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=3,
+        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.)
+        ]) * 3),
+    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=[(15, 15), (11, 11), (9, 9), (7, 7)] * 2 + [(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/3xrsn50_coco_256x192.py

+_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=3,
+        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=3,
+        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.)
+        ]) * 3),
+    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=[(15, 15), (11, 11), (9, 9), (7, 7)] * 2 + [(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/4xmspn50_coco_256x192.py

+_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='torchvision://resnet50',
+    backbone=dict(
+        type='MSPN',
+        unit_channels=256,
+        num_stages=4,
+        num_units=4,
+        num_blocks=[3, 4, 6, 3],
+        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=4,
+        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.)
+        ]) * 4),
+    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=[(15, 15), (11, 11), (9, 9), (7, 7)] * 3 + [(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/alexnet_coco.md

+<!-- [BACKBONE] -->
+
+<details>
+<summary align="right"><a href="https://proceedings.neurips.cc/paper/4824-imagenet-classification-with-deep-convolutional-neural-networks.pdf">AlexNet (NeurIPS'2012)</a></summary>
+
+```bibtex
+@inproceedings{krizhevsky2012imagenet,
+  title={Imagenet classification with deep convolutional neural networks},
+  author={Krizhevsky, Alex and Sutskever, Ilya and Hinton, Geoffrey E},
+  booktitle={Advances in neural information processing systems},
+  pages={1097--1105},
+  year={2012}
+}
+```
+
+</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_alexnet](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/alexnet_coco_256x192.py) |  256x192   | 0.397 |      0.758      |      0.381      | 0.478 |      0.822      | [ckpt](https://download.openmmlab.com/mmpose/top_down/alexnet/alexnet_coco_256x192-a7b1fd15_20200727.pth) | [log](https://download.openmmlab.com/mmpose/top_down/alexnet/alexnet_coco_256x192_20200727.log.json) |

configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/alexnet_coco.yml

+Collections:
+- Name: AlexNet
+  Paper:
+    Title: Imagenet classification with deep convolutional neural networks
+    URL: https://proceedings.neurips.cc/paper/4824-imagenet-classification-with-deep-convolutional-neural-networks.pdf
+  README: https://github.com/open-mmlab/mmpose/blob/master/docs/en/papers/backbones/alexnet.md
+Models:
+- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/alexnet_coco_256x192.py
+  In Collection: AlexNet
+  Metadata:
+    Architecture:
+    - AlexNet
+    Training Data: COCO
+  Name: topdown_heatmap_alexnet_coco_256x192
+  Results:
+  - Dataset: COCO
+    Metrics:
+      AP: 0.397
+      AP@0.5: 0.758
+      AP@0.75: 0.381
+      AR: 0.478
+      AR@0.5: 0.822
+    Task: Body 2D Keypoint
+  Weights: https://download.openmmlab.com/mmpose/top_down/alexnet/alexnet_coco_256x192-a7b1fd15_20200727.pth

configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/alexnet_coco_256x192.py

+_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=None,
+    backbone=dict(type='AlexNet', num_classes=-1),
+    keypoint_head=dict(
+        type='TopdownHeatmapSimpleHead',
+        in_channels=256,
+        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=[40, 56],
+    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/cpm_coco.md

+<!-- [ALGORITHM] -->
+
+<details>
+<summary align="right"><a href="http://openaccess.thecvf.com/content_cvpr_2016/html/Wei_Convolutional_Pose_Machines_CVPR_2016_paper.html">CPM (CVPR'2016)</a></summary>
+
+```bibtex
+@inproceedings{wei2016convolutional,
+  title={Convolutional pose machines},
+  author={Wei, Shih-En and Ramakrishna, Varun and Kanade, Takeo and Sheikh, Yaser},
+  booktitle={Proceedings of the IEEE conference on Computer Vision and Pattern Recognition},
+  pages={4724--4732},
+  year={2016}
+}
+```
+
+</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                      |
+| :-------------------------------------------- | :--------: | :---: | :-------------: | :-------------: | :---: | :-------------: | :-------------------------------------------: | :-------------------------------------------: |
+| [cpm](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/cpm_coco_256x192.py) |  256x192   | 0.623 |      0.859      |      0.704      | 0.686 |      0.903      | [ckpt](https://download.openmmlab.com/mmpose/top_down/cpm/cpm_coco_256x192-aa4ba095_20200817.pth) | [log](https://download.openmmlab.com/mmpose/top_down/cpm/cpm_coco_256x192_20200817.log.json) |
+| [cpm](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/cpm_coco_384x288.py) |  384x288   | 0.650 |      0.864      |      0.725      | 0.708 |      0.905      | [ckpt](https://download.openmmlab.com/mmpose/top_down/cpm/cpm_coco_384x288-80feb4bc_20200821.pth) | [log](https://download.openmmlab.com/mmpose/top_down/cpm/cpm_coco_384x288_20200821.log.json) |

configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/cpm_coco.yml

+Collections:
+- Name: CPM
+  Paper:
+    Title: Convolutional pose machines
+    URL: http://openaccess.thecvf.com/content_cvpr_2016/html/Wei_Convolutional_Pose_Machines_CVPR_2016_paper.html
+  README: https://github.com/open-mmlab/mmpose/blob/master/docs/en/papers/backbones/cpm.md
+Models:
+- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/cpm_coco_256x192.py
+  In Collection: CPM
+  Metadata:
+    Architecture: &id001
+    - CPM
+    Training Data: COCO
+  Name: topdown_heatmap_cpm_coco_256x192
+  Results:
+  - Dataset: COCO
+    Metrics:
+      AP: 0.623
+      AP@0.5: 0.859
+      AP@0.75: 0.704
+      AR: 0.686
+      AR@0.5: 0.903
+    Task: Body 2D Keypoint
+  Weights: https://download.openmmlab.com/mmpose/top_down/cpm/cpm_coco_256x192-aa4ba095_20200817.pth
+- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/cpm_coco_384x288.py
+  In Collection: CPM
+  Metadata:
+    Architecture: *id001
+    Training Data: COCO
+  Name: topdown_heatmap_cpm_coco_384x288
+  Results:
+  - Dataset: COCO
+    Metrics:
+      AP: 0.65
+      AP@0.5: 0.864
+      AP@0.75: 0.725
+      AR: 0.708
+      AR@0.5: 0.905
+    Task: Body 2D Keypoint
+  Weights: https://download.openmmlab.com/mmpose/top_down/cpm/cpm_coco_384x288-80feb4bc_20200821.pth

configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/cpm_coco_256x192.py

+_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=None,
+    backbone=dict(
+        type='CPM',
+        in_channels=3,
+        out_channels=channel_cfg['num_output_channels'],
+        feat_channels=128,
+        num_stages=6),
+    keypoint_head=dict(
+        type='TopdownHeatmapMultiStageHead',
+        in_channels=channel_cfg['num_output_channels'],
+        out_channels=channel_cfg['num_output_channels'],
+        num_stages=6,
+        num_deconv_layers=0,
+        extra=dict(final_conv_kernel=0, ),
+        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=[24, 32],
+    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/cpm_coco_384x288.py

+_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=None,
+    backbone=dict(
+        type='CPM',
+        in_channels=3,
+        out_channels=channel_cfg['num_output_channels'],
+        feat_channels=128,
+        num_stages=6),
+    keypoint_head=dict(
+        type='TopdownHeatmapMultiStageHead',
+        in_channels=channel_cfg['num_output_channels'],
+        out_channels=channel_cfg['num_output_channels'],
+        num_stages=6,
+        num_deconv_layers=0,
+        extra=dict(final_conv_kernel=0, ),
+        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=[36, 48],
+    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/hourglass52_coco_256x256.py

+_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=None,
+    backbone=dict(
+        type='HourglassNet',
+        num_stacks=1,
+    ),
+    keypoint_head=dict(
+        type='TopdownHeatmapMultiStageHead',
+        in_channels=256,
+        out_channels=channel_cfg['num_output_channels'],
+        num_stages=1,
+        num_deconv_layers=0,
+        extra=dict(final_conv_kernel=1, ),
+        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=[256, 256],
+    heatmap_size=[64, 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/hourglass52_coco_384x384.py

+_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=None,
+    backbone=dict(
+        type='HourglassNet',
+        num_stacks=1,
+    ),
+    keypoint_head=dict(
+        type='TopdownHeatmapMultiStageHead',
+        in_channels=256,
+        out_channels=channel_cfg['num_output_channels'],
+        num_stages=1,
+        num_deconv_layers=0,
+        extra=dict(final_conv_kernel=1, ),
+        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=[384, 384],
+    heatmap_size=[96, 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/hourglass_coco.md

+<!-- [ALGORITHM] -->
+
+<details>
+<summary align="right"><a href="https://link.springer.com/chapter/10.1007/978-3-319-46484-8_29">Hourglass (ECCV'2016)</a></summary>
+
+```bibtex
+@inproceedings{newell2016stacked,
+  title={Stacked hourglass networks for human pose estimation},
+  author={Newell, Alejandro and Yang, Kaiyu and Deng, Jia},
+  booktitle={European conference on computer vision},
+  pages={483--499},
+  year={2016},
+  organization={Springer}
+}
+```
+
+</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_hourglass_52](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hourglass52_coco_256x256.py) |  256x256   | 0.726 |      0.896      |      0.799      | 0.780 |      0.934      | [ckpt](https://download.openmmlab.com/mmpose/top_down/hourglass/hourglass52_coco_256x256-4ec713ba_20200709.pth) | [log](https://download.openmmlab.com/mmpose/top_down/hourglass/hourglass52_coco_256x256_20200709.log.json) |
+| [pose_hourglass_52](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hourglass52_coco_384x384.py) |  384x384   | 0.746 |      0.900      |      0.813      | 0.797 |      0.939      | [ckpt](https://download.openmmlab.com/mmpose/top_down/hourglass/hourglass52_coco_384x384-be91ba2b_20200812.pth) | [log](https://download.openmmlab.com/mmpose/top_down/hourglass/hourglass52_coco_384x384_20200812.log.json) |

configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hourglass_coco.yml

+Collections:
+- Name: Hourglass
+  Paper:
+    Title: Stacked hourglass networks for human pose estimation
+    URL: https://link.springer.com/chapter/10.1007/978-3-319-46484-8_29
+  README: https://github.com/open-mmlab/mmpose/blob/master/docs/en/papers/backbones/hourglass.md
+Models:
+- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hourglass52_coco_256x256.py
+  In Collection: Hourglass
+  Metadata:
+    Architecture: &id001
+    - Hourglass
+    Training Data: COCO
+  Name: topdown_heatmap_hourglass52_coco_256x256
+  Results:
+  - Dataset: COCO
+    Metrics:
+      AP: 0.726
+      AP@0.5: 0.896
+      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/hourglass/hourglass52_coco_256x256-4ec713ba_20200709.pth
+- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hourglass52_coco_384x384.py
+  In Collection: Hourglass
+  Metadata:
+    Architecture: *id001
+    Training Data: COCO
+  Name: topdown_heatmap_hourglass52_coco_384x384
+  Results:
+  - Dataset: COCO
+    Metrics:
+      AP: 0.746
+      AP@0.5: 0.9
+      AP@0.75: 0.813
+      AR: 0.797
+      AR@0.5: 0.939
+    Task: Body 2D Keypoint
+  Weights: https://download.openmmlab.com/mmpose/top_down/hourglass/hourglass52_coco_384x384-be91ba2b_20200812.pth

configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hrformer_base_coco_256x192.py

+_base_ = [
+    '../../../../_base_/default_runtime.py',
+    '../../../../_base_/datasets/coco.py'
+]
+
+evaluation = dict(interval=10, metric='mAP', save_best='AP')
+
+optimizer = dict(
+    type='AdamW',
+    lr=5e-4,
+    betas=(0.9, 0.999),
+    weight_decay=0.01,
+    paramwise_cfg=dict(
+        custom_keys={'relative_position_bias_table': dict(decay_mult=0.)}))
+
+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
+norm_cfg = dict(type='SyncBN', requires_grad=True)
+model = dict(
+    type='TopDown',
+    pretrained='https://download.openmmlab.com/mmpose/'
+    'pretrain_models/hrformer_base-32815020_20220226.pth',
+    backbone=dict(
+        type='HRFormer',
+        in_channels=3,
+        norm_cfg=norm_cfg,
+        extra=dict(
+            drop_path_rate=0.2,
+            with_rpe=True,
+            stage1=dict(
+                num_modules=1,
+                num_branches=1,
+                block='BOTTLENECK',
+                num_blocks=(2, ),
+                num_channels=(64, ),
+                num_heads=[2],
+                mlp_ratios=[4]),
+            stage2=dict(
+                num_modules=1,
+                num_branches=2,
+                block='HRFORMERBLOCK',
+                num_blocks=(2, 2),
+                num_channels=(78, 156),
+                num_heads=[2, 4],
+                mlp_ratios=[4, 4],
+                window_sizes=[7, 7]),
+            stage3=dict(
+                num_modules=4,
+                num_branches=3,
+                block='HRFORMERBLOCK',
+                num_blocks=(2, 2, 2),
+                num_channels=(78, 156, 312),
+                num_heads=[2, 4, 8],
+                mlp_ratios=[4, 4, 4],
+                window_sizes=[7, 7, 7]),
+            stage4=dict(
+                num_modules=2,
+                num_branches=4,
+                block='HRFORMERBLOCK',
+                num_blocks=(2, 2, 2, 2),
+                num_channels=(78, 156, 312, 624),
+                num_heads=[2, 4, 8, 16],
+                mlp_ratios=[4, 4, 4, 4],
+                window_sizes=[7, 7, 7, 7]))),
+    keypoint_head=dict(
+        type='TopdownHeatmapSimpleHead',
+        in_channels=78,
+        out_channels=channel_cfg['num_output_channels'],
+        num_deconv_layers=0,
+        extra=dict(final_conv_kernel=1, ),
+        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_root = 'data/coco'
+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=f'{data_root}/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 = 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=val_pipeline,
+        dataset_info={{_base_.dataset_info}}),
+)
+
+# fp16 settings
+fp16 = dict(loss_scale='dynamic')

configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hrformer_base_coco_384x288.py

+_base_ = [
+    '../../../../_base_/default_runtime.py',
+    '../../../../_base_/datasets/coco.py'
+]
+
+evaluation = dict(interval=10, metric='mAP', save_best='AP')
+
+optimizer = dict(
+    type='AdamW',
+    lr=5e-4,
+    betas=(0.9, 0.999),
+    weight_decay=0.01,
+    paramwise_cfg=dict(
+        custom_keys={'relative_position_bias_table': dict(decay_mult=0.)}))
+
+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
+norm_cfg = dict(type='SyncBN', requires_grad=True)
+model = dict(
+    type='TopDown',
+    pretrained='https://download.openmmlab.com/mmpose/'
+    'pretrain_models/hrformer_base-32815020_20220226.pth',
+    backbone=dict(
+        type='HRFormer',
+        in_channels=3,
+        norm_cfg=norm_cfg,
+        extra=dict(
+            drop_path_rate=0.3,
+            with_rpe=True,
+            stage1=dict(
+                num_modules=1,
+                num_branches=1,
+                block='BOTTLENECK',
+                num_blocks=(2, ),
+                num_channels=(64, ),
+                num_heads=[2],
+                mlp_ratios=[4]),
+            stage2=dict(
+                num_modules=1,
+                num_branches=2,
+                block='HRFORMERBLOCK',
+                num_blocks=(2, 2),
+                num_channels=(78, 156),
+                num_heads=[2, 4],
+                mlp_ratios=[4, 4],
+                window_sizes=[7, 7]),
+            stage3=dict(
+                num_modules=4,
+                num_branches=3,
+                block='HRFORMERBLOCK',
+                num_blocks=(2, 2, 2),
+                num_channels=(78, 156, 312),
+                num_heads=[2, 4, 8],
+                mlp_ratios=[4, 4, 4],
+                window_sizes=[7, 7, 7]),
+            stage4=dict(
+                num_modules=2,
+                num_branches=4,
+                block='HRFORMERBLOCK',
+                num_blocks=(2, 2, 2, 2),
+                num_channels=(78, 156, 312, 624),
+                num_heads=[2, 4, 8, 16],
+                mlp_ratios=[4, 4, 4, 4],
+                window_sizes=[7, 7, 7, 7]))),
+    keypoint_head=dict(
+        type='TopdownHeatmapSimpleHead',
+        in_channels=78,
+        out_channels=channel_cfg['num_output_channels'],
+        num_deconv_layers=0,
+        extra=dict(final_conv_kernel=1, ),
+        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=17))
+
+data_root = 'data/coco'
+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=f'{data_root}/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 = 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=val_pipeline,
+        dataset_info={{_base_.dataset_info}}),
+)
+
+# fp16 settings
+fp16 = dict(loss_scale='dynamic')

configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hrformer_coco.md

+<!-- [ALGORITHM] -->
+
+<details>
+<summary align="right"><a href="https://proceedings.neurips.cc/paper/2021/hash/3bbfdde8842a5c44a0323518eec97cbe-Abstract.html">HRFormer (NIPS'2021)</a></summary>
+
+```bibtex
+@article{yuan2021hrformer,
+  title={HRFormer: High-Resolution Vision Transformer for Dense Predict},
+  author={Yuan, Yuhui and Fu, Rao and Huang, Lang and Lin, Weihong and Zhang, Chao and Chen, Xilin and Wang, Jingdong},
+  journal={Advances in Neural Information Processing Systems},
+  volume={34},
+  year={2021}
+}
+```
+
+</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_hrformer_small](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hrformer_small_coco_256x192.py) |  256x192   | 0.738 |      0.904      |      0.811      | 0.792 |      0.941      | [ckpt](https://download.openmmlab.com/mmpose/top_down/hrformer/hrformer_small_coco_256x192-5310d898_20220316.pth) | [log](https://download.openmmlab.com/mmpose/top_down/hrformer/hrformer_small_coco_256x192_20220316.log.json) |
+| [pose_hrformer_small](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hrformer_small_coco_384x288.py) |  384x288   | 0.757 |      0.905      |      0.824      | 0.807 |      0.941      | [ckpt](https://download.openmmlab.com/mmpose/top_down/hrformer/hrformer_small_coco_384x288-98d237ed_20220316.pth) | [log](https://download.openmmlab.com/mmpose/top_down/hrformer/hrformer_small_coco_384x288_20220316.log.json) |
+| [pose_hrformer_base](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hrformer_base_coco_256x192.py) |  256x192   | 0.753 |      0.907      |      0.826      | 0.807 |      0.943      | [ckpt](https://download.openmmlab.com/mmpose/top_down/hrformer/hrformer_base_coco_256x192-6f5f1169_20220316.pth) | [log](https://download.openmmlab.com/mmpose/top_down/hrformer/hrformer_base_coco_256x192_20220316.log.json) |
+| [pose_hrformer_base](/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hrformer_base_coco_384x288.py) |  384x288   | 0.774 |      0.909      |      0.842      | 0.823 |      0.945      | [ckpt](https://download.openmmlab.com/mmpose/top_down/hrformer/hrformer_base_coco_384x288-ecf0758d_20220316.pth) | [log](https://download.openmmlab.com/mmpose/top_down/hrformer/hrformer_base_coco_256x192_20220316.log.json) |

configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hrformer_coco.yml

+Collections:
+- Name: HRFormer
+  Paper:
+    Title: 'HRFormer: High-Resolution Vision Transformer for Dense Predict'
+    URL: https://proceedings.neurips.cc/paper/2021/hash/3bbfdde8842a5c44a0323518eec97cbe-Abstract.html
+  README: https://github.com/open-mmlab/mmpose/blob/master/docs/en/papers/backbones/hrformer.md
+Models:
+- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hrformer_small_coco_256x192.py
+  In Collection: HRFormer
+  Metadata:
+    Architecture: &id001
+    - HRFormer
+    Training Data: COCO
+  Name: topdown_heatmap_hrformer_small_coco_256x192
+  Results:
+  - Dataset: COCO
+    Metrics:
+      AP: 0.738
+      AP@0.5: 0.904
+      AP@0.75: 0.811
+      AR: 0.792
+      AR@0.5: 0.941
+    Task: Body 2D Keypoint
+  Weights: https://download.openmmlab.com/mmpose/top_down/hrformer/hrformer_small_coco_256x192-5310d898_20220316.pth
+- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hrformer_small_coco_384x288.py
+  In Collection: HRFormer
+  Metadata:
+    Architecture: *id001
+    Training Data: COCO
+  Name: topdown_heatmap_hrformer_small_coco_384x288
+  Results:
+  - Dataset: COCO
+    Metrics:
+      AP: 0.757
+      AP@0.5: 0.905
+      AP@0.75: 0.824
+      AR: 0.807
+      AR@0.5: 0.941
+    Task: Body 2D Keypoint
+  Weights: https://download.openmmlab.com/mmpose/top_down/hrformer/hrformer_small_coco_384x288-98d237ed_20220316.pth
+- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hrformer_base_coco_256x192.py
+  In Collection: HRFormer
+  Metadata:
+    Architecture: *id001
+    Training Data: COCO
+  Name: topdown_heatmap_hrformer_base_coco_256x192
+  Results:
+  - Dataset: COCO
+    Metrics:
+      AP: 0.753
+      AP@0.5: 0.907
+      AP@0.75: 0.826
+      AR: 0.807
+      AR@0.5: 0.943
+    Task: Body 2D Keypoint
+  Weights: https://download.openmmlab.com/mmpose/top_down/hrformer/hrformer_base_coco_256x192-6f5f1169_20220316.pth
+- Config: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hrformer_base_coco_384x288.py
+  In Collection: HRFormer
+  Metadata:
+    Architecture: *id001
+    Training Data: COCO
+  Name: topdown_heatmap_hrformer_base_coco_384x288
+  Results:
+  - Dataset: COCO
+    Metrics:
+      AP: 0.774
+      AP@0.5: 0.909
+      AP@0.75: 0.842
+      AR: 0.823
+      AR@0.5: 0.945
+    Task: Body 2D Keypoint
+  Weights: https://download.openmmlab.com/mmpose/top_down/hrformer/hrformer_base_coco_384x288-ecf0758d_20220316.pth

configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hrformer_small_coco_256x192.py

+_base_ = [
+    '../../../../_base_/default_runtime.py',
+    '../../../../_base_/datasets/coco.py'
+]
+
+evaluation = dict(interval=10, metric='mAP', save_best='AP')
+
+optimizer = dict(
+    type='AdamW',
+    lr=5e-4,
+    betas=(0.9, 0.999),
+    weight_decay=0.01,
+    paramwise_cfg=dict(
+        custom_keys={'relative_position_bias_table': dict(decay_mult=0.)}))
+
+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
+norm_cfg = dict(type='SyncBN', requires_grad=True)
+model = dict(
+    type='TopDown',
+    pretrained='https://download.openmmlab.com/mmpose/'
+    'pretrain_models/hrformer_small-09516375_20220226.pth',
+    backbone=dict(
+        type='HRFormer',
+        in_channels=3,
+        norm_cfg=norm_cfg,
+        extra=dict(
+            drop_path_rate=0.1,
+            with_rpe=True,
+            stage1=dict(
+                num_modules=1,
+                num_branches=1,
+                block='BOTTLENECK',
+                num_blocks=(2, ),
+                num_channels=(64, ),
+                num_heads=[2],
+                num_mlp_ratios=[4]),
+            stage2=dict(
+                num_modules=1,
+                num_branches=2,
+                block='HRFORMERBLOCK',
+                num_blocks=(2, 2),
+                num_channels=(32, 64),
+                num_heads=[1, 2],
+                mlp_ratios=[4, 4],
+                window_sizes=[7, 7]),
+            stage3=dict(
+                num_modules=4,
+                num_branches=3,
+                block='HRFORMERBLOCK',
+                num_blocks=(2, 2, 2),
+                num_channels=(32, 64, 128),
+                num_heads=[1, 2, 4],
+                mlp_ratios=[4, 4, 4],
+                window_sizes=[7, 7, 7]),
+            stage4=dict(
+                num_modules=2,
+                num_branches=4,
+                block='HRFORMERBLOCK',
+                num_blocks=(2, 2, 2, 2),
+                num_channels=(32, 64, 128, 256),
+                num_heads=[1, 2, 4, 8],
+                mlp_ratios=[4, 4, 4, 4],
+                window_sizes=[7, 7, 7, 7]))),
+    keypoint_head=dict(
+        type='TopdownHeatmapSimpleHead',
+        in_channels=32,
+        out_channels=channel_cfg['num_output_channels'],
+        num_deconv_layers=0,
+        extra=dict(final_conv_kernel=1, ),
+        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_root = 'data/coco'
+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=f'{data_root}/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 = 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=val_pipeline,
+        dataset_info={{_base_.dataset_info}}),
+)
+
+# fp16 settings
+fp16 = dict(loss_scale='dynamic')

configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hrformer_small_coco_384x288.py

+_base_ = [
+    '../../../../_base_/default_runtime.py',
+    '../../../../_base_/datasets/coco.py'
+]
+
+evaluation = dict(interval=10, metric='mAP', save_best='AP')
+
+optimizer = dict(
+    type='AdamW',
+    lr=5e-4,
+    betas=(0.9, 0.999),
+    weight_decay=0.01,
+    paramwise_cfg=dict(
+        custom_keys={'relative_position_bias_table': dict(decay_mult=0.)}))
+
+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
+norm_cfg = dict(type='SyncBN', requires_grad=True)
+model = dict(
+    type='TopDown',
+    pretrained='https://download.openmmlab.com/mmpose/'
+    'pretrain_models/hrformer_small-09516375_20220226.pth',
+    backbone=dict(
+        type='HRFormer',
+        in_channels=3,
+        norm_cfg=norm_cfg,
+        extra=dict(
+            drop_path_rate=0.1,
+            with_rpe=True,
+            stage1=dict(
+                num_modules=1,
+                num_branches=1,
+                block='BOTTLENECK',
+                num_blocks=(2, ),
+                num_channels=(64, ),
+                num_heads=[2],
+                num_mlp_ratios=[4]),
+            stage2=dict(
+                num_modules=1,
+                num_branches=2,
+                block='HRFORMERBLOCK',
+                num_blocks=(2, 2),
+                num_channels=(32, 64),
+                num_heads=[1, 2],
+                mlp_ratios=[4, 4],
+                window_sizes=[7, 7]),
+            stage3=dict(
+                num_modules=4,
+                num_branches=3,
+                block='HRFORMERBLOCK',
+                num_blocks=(2, 2, 2),
+                num_channels=(32, 64, 128),
+                num_heads=[1, 2, 4],
+                mlp_ratios=[4, 4, 4],
+                window_sizes=[7, 7, 7]),
+            stage4=dict(
+                num_modules=2,
+                num_branches=4,
+                block='HRFORMERBLOCK',
+                num_blocks=(2, 2, 2, 2),
+                num_channels=(32, 64, 128, 256),
+                num_heads=[1, 2, 4, 8],
+                mlp_ratios=[4, 4, 4, 4],
+                window_sizes=[7, 7, 7, 7]))),
+    keypoint_head=dict(
+        type='TopdownHeatmapSimpleHead',
+        in_channels=32,
+        out_channels=channel_cfg['num_output_channels'],
+        num_deconv_layers=0,
+        extra=dict(final_conv_kernel=1, ),
+        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_root = 'data/coco'
+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=f'{data_root}/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 = 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=val_pipeline,
+        dataset_info={{_base_.dataset_info}}),
+)
+
+# fp16 settings
+fp16 = dict(loss_scale='dynamic')