Merge branch 'dev-1.x' into interhand26m

LareinaM · Sep 15, 2023 · 8fcaca5 · 8fcaca5
2 parents b1d51e3 + 5530c3b
commit 8fcaca5
Show file tree

Hide file tree

Showing 96 changed files with 6,250 additions and 120 deletions.
diff --git a/...ody_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-base-simple_8xb64-210e_coco-256x192.py b/...ody_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-base-simple_8xb64-210e_coco-256x192.py
@@ -71,7 +71,7 @@
         init_cfg=dict(
             type='Pretrained',
             checkpoint='https://download.openmmlab.com/mmpose/'
-            'v1/pretrained_models/mae_pretrain_vit_base.pth'),
+            'v1/pretrained_models/mae_pretrain_vit_base_20230913.pth'),
     ),
     neck=dict(type='FeatureMapProcessor', scale_factor=4.0, apply_relu=True),
     head=dict(

diff --git a/configs/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-base_8xb64-210e_coco-256x192.py b/configs/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-base_8xb64-210e_coco-256x192.py
@@ -71,7 +71,7 @@
         init_cfg=dict(
             type='Pretrained',
             checkpoint='https://download.openmmlab.com/mmpose/'
-            'v1/pretrained_models/mae_pretrain_vit_base.pth'),
+            'v1/pretrained_models/mae_pretrain_vit_base_20230913.pth'),
     ),
     head=dict(
         type='HeatmapHead',

diff --git a/...ody_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-huge-simple_8xb64-210e_coco-256x192.py b/...ody_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-huge-simple_8xb64-210e_coco-256x192.py
@@ -71,7 +71,7 @@
         init_cfg=dict(
             type='Pretrained',
             checkpoint='https://download.openmmlab.com/mmpose/'
-            'v1/pretrained_models/mae_pretrain_vit_huge.pth'),
+            'v1/pretrained_models/mae_pretrain_vit_huge_20230913.pth'),
     ),
     neck=dict(type='FeatureMapProcessor', scale_factor=4.0, apply_relu=True),
     head=dict(

diff --git a/configs/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-huge_8xb64-210e_coco-256x192.py b/configs/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-huge_8xb64-210e_coco-256x192.py
@@ -71,7 +71,7 @@
         init_cfg=dict(
             type='Pretrained',
             checkpoint='https://download.openmmlab.com/mmpose/'
-            'v1/pretrained_models/mae_pretrain_vit_huge.pth'),
+            'v1/pretrained_models/mae_pretrain_vit_huge_20230913.pth'),
     ),
     head=dict(
         type='HeatmapHead',

diff --git a/...dy_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-large-simple_8xb64-210e_coco-256x192.py b/...dy_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-large-simple_8xb64-210e_coco-256x192.py
@@ -71,7 +71,7 @@
         init_cfg=dict(
             type='Pretrained',
             checkpoint='https://download.openmmlab.com/mmpose/'
-            'v1/pretrained_models/mae_pretrain_vit_large.pth'),
+            'v1/pretrained_models/mae_pretrain_vit_large_20230913.pth'),
     ),
     neck=dict(type='FeatureMapProcessor', scale_factor=4.0, apply_relu=True),
     head=dict(

diff --git a/configs/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-large_8xb64-210e_coco-256x192.py b/configs/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-large_8xb64-210e_coco-256x192.py
@@ -71,7 +71,7 @@
         init_cfg=dict(
             type='Pretrained',
             checkpoint='https://download.openmmlab.com/mmpose/'
-            'v1/pretrained_models/mae_pretrain_vit_large.pth'),
+            'v1/pretrained_models/mae_pretrain_vit_large_20230913.pth'),
     ),
     head=dict(
         type='HeatmapHead',

diff --git a/...dy_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-small-simple_8xb64-210e_coco-256x192.py b/...dy_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-small-simple_8xb64-210e_coco-256x192.py
@@ -76,7 +76,7 @@
         init_cfg=dict(
             type='Pretrained',
             checkpoint='https://download.openmmlab.com/mmpose/'
-            'v1/pretrained_models/mae_pretrain_vit_small.pth'),
+            'v1/pretrained_models/mae_pretrain_vit_small_20230913.pth'),
     ),
     neck=dict(type='FeatureMapProcessor', scale_factor=4.0, apply_relu=True),
     head=dict(

diff --git a/configs/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-small_8xb64-210e_coco-256x192.py b/configs/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-small_8xb64-210e_coco-256x192.py
@@ -76,7 +76,7 @@
         init_cfg=dict(
             type='Pretrained',
             checkpoint='https://download.openmmlab.com/mmpose/'
-            'v1/pretrained_models/mae_pretrain_vit_small.pth'),
+            'v1/pretrained_models/mae_pretrain_vit_small_20230913.pth'),
     ),
     head=dict(
         type='HeatmapHead',

diff --git a/configs/body_2d_keypoint/yoloxpose/README.md b/configs/body_2d_keypoint/yoloxpose/README.md
@@ -0,0 +1,22 @@
+# YOLO-Pose: Enhancing YOLO for Multi Person Pose Estimation Using Object Keypoint Similarity Loss
+
+<!-- [ALGORITHM] -->
+
+<details>
+<summary align="right"><a href="https://arxiv.org/abs/2204.06806">YOLO-Pose (CVPRW'2022)</a></summary>
+
+```bibtex
+@inproceedings{maji2022yolo,
+  title={Yolo-pose: Enhancing yolo for multi person pose estimation using object keypoint similarity loss},
+  author={Maji, Debapriya and Nagori, Soyeb and Mathew, Manu and Poddar, Deepak},
+  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
+  pages={2637--2646},
+  year={2022}
+}
+```
+
+</details>
+
+YOLO-Pose is a bottom-up pose estimation approach that simultaneously detects all person instances and regresses keypoint locations in a single pass.
+
+We implement **YOLOX-Pose** based on the **YOLOX** object detection framework and inherits the benefits of unified pose estimation and object detection from YOLO-pose. To predict keypoint locations more accurately, separate branches with adaptive convolutions are used to regress the offsets for different joints. This allows optimizing the feature extraction for each keypoint.
diff --git a/configs/body_2d_keypoint/yoloxpose/coco/yoloxpose_coco.md b/configs/body_2d_keypoint/yoloxpose/coco/yoloxpose_coco.md
@@ -0,0 +1,59 @@
+<!-- [ALGORITHM] -->
+
+<details>
+<summary align="right"><a href="https://arxiv.org/abs/2204.06806">YOLO-Pose (CVPRW'2022)</a></summary>
+
+```bibtex
+@inproceedings{maji2022yolo,
+  title={Yolo-pose: Enhancing yolo for multi person pose estimation using object keypoint similarity loss},
+  author={Maji, Debapriya and Nagori, Soyeb and Mathew, Manu and Poddar, Deepak},
+  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
+  pages={2637--2646},
+  year={2022}
+}
+```
+
+</details>
+
+<!-- [BACKBONE] -->
+
+<details>
+<summary align="right"><a href="https://arxiv.org/abs/2107.08430">YOLOX</a></summary>
+
+```bibtex
+@article{ge2021yolox,
+  title={Yolox: Exceeding yolo series in 2021},
+  author={Ge, Zheng and Liu, Songtao and Wang, Feng and Li, Zeming and Sun, Jian},
+  journal={arXiv preprint arXiv:2107.08430},
+  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                      |
+| :-------------------------------------------- | :--------: | :---: | :-------------: | :-------------: | :---: | :-------------: | :-------------------------------------------: | :-------------------------------------------: |
+| [yoloxpose_tiny](/configs/body_2d_keypoint/yoloxpose/coco/yoloxpose_tiny_4xb64-300e_coco-416.py) |  416x416   | 0.527 |      0.794      |      0.557      | 0.577 |      0.843      | [ckpt](https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/yolox_pose/yoloxpose_tiny_4xb64-300e_coco-416-76eb44ca_20230829.pth) | [log](https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/yolox_pose/yoloxpose_tiny_4xb64-300e_coco-416-20230829.json) |
+| [yoloxpose_s](/configs/body_2d_keypoint/yoloxpose/coco/yoloxpose_s_8xb32-300e_coco-640.py) |  640x640   | 0.642 |      0.873      |      0.702      | 0.688 |      0.912      | [ckpt](https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/yolox_pose/yoloxpose_s_8xb32-300e_coco-640-56c79c1f_20230829.pth) | [log](https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/yolox_pose/yoloxpose_s_8xb32-300e_coco-640-20230829.json) |
+| [yoloxpose_m](/configs/body_2d_keypoint/yoloxpose/coco/yoloxpose_m_8xb32-300e_coco-640.py) |  640x640   | 0.697 |      0.903      |      0.766      | 0.739 |      0.933      | [ckpt](https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/yolox_pose/yoloxpose_m_8xb32-300e_coco-640-84e9a538_20230829.pth) | [log](https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/yolox_pose/yoloxpose_m_8xb32-300e_coco-640-20230829.json) |
+| [yoloxpose_l](/configs/body_2d_keypoint/yoloxpose/coco/yoloxpose_l_8xb32-300e_coco-640.py) |  640x640   | 0.714 |      0.906      |      0.785      | 0.756 |      0.934      | [ckpt](https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/yolox_pose/yoloxpose_l_8xb32-300e_coco-640-de0f8dee_20230829.pth) | [log](https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/yolox_pose/yoloxpose_l_8xb32-300e_coco-640-20230829.json) |
diff --git a/configs/body_2d_keypoint/yoloxpose/coco/yoloxpose_coco.yml b/configs/body_2d_keypoint/yoloxpose/coco/yoloxpose_coco.yml
@@ -0,0 +1,72 @@
+Collections:
+- Name: YOLOXPose
+  Paper:
+    Title: 'YOLO-Pose: Enhancing YOLO for Multi Person Pose Estimation Using Object Keypoint Similarity Loss'
+    URL: https://arxiv.org/abs/2204.06806
+  README: https://github.com/open-mmlab/mmpose/blob/main/docs/src/papers/algorithms/yolopose.md
+Models:
+- Config: configs/body_2d_keypoint/yoloxpose/coco/yoloxpose_tiny_4xb64-300e_coco-416.py
+  In Collection: YOLOXPose
+  Metadata:
+    Architecture: &id001
+    - YOLOXPose
+    Training Data: COCO
+  Name: yoloxpose_tiny_4xb64-300e_coco-416
+  Results:
+  - Dataset: COCO
+    Metrics:
+      AP: 0.527
+      [email protected]: 0.794
+      [email protected]: 0.557
+      AR: 0.577
+      [email protected]: 0.843
+    Task: Body 2D Keypoint
+  Weights: https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/yolox_pose/yoloxpose_tiny_4xb64-300e_coco-416-76eb44ca_20230829.pth
+- Config: configs/body_2d_keypoint/yoloxpose/coco/yoloxpose_s_8xb32-300e_coco-640.py
+  In Collection: YOLOXPose
+  Metadata:
+    Architecture: *id001
+    Training Data: COCO
+  Name: yoloxpose_s_8xb32-300e_coco-640
+  Results:
+  - Dataset: COCO
+    Metrics:
+      AP: 0.642
+      [email protected]: 0.873
+      [email protected]: 0.702
+      AR: 0.688
+      [email protected]: 0.912
+    Task: Body 2D Keypoint
+  Weights: https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/yolox_pose/yoloxpose_s_8xb32-300e_coco-640-56c79c1f_20230829.pth
+- Config: configs/body_2d_keypoint/yoloxpose/coco/yoloxpose_m_8xb32-300e_coco-640.py
+  In Collection: YOLOXPose
+  Metadata:
+    Architecture: *id001
+    Training Data: COCO
+  Name: yoloxpose_m_8xb32-300e_coco-640
+  Results:
+  - Dataset: COCO
+    Metrics:
+      AP: 0.697
+      [email protected]: 0.903
+      [email protected]: 0.766
+      AR: 0.739
+      [email protected]: 0.933
+    Task: Body 2D Keypoint
+  Weights: https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/yolox_pose/yoloxpose_m_8xb32-300e_coco-640-84e9a538_20230829.pth
+- Config: configs/body_2d_keypoint/yoloxpose/coco/yoloxpose_l_8xb32-300e_coco-640.py
+  In Collection: YOLOXPose
+  Metadata:
+    Architecture: *id001
+    Training Data: COCO
+  Name: yoloxpose_l_8xb32-300e_coco-640
+  Results:
+  - Dataset: COCO
+    Metrics:
+      AP: 0.714
+      [email protected]: 0.906
+      [email protected]: 0.785
+      AR: 0.756
+      [email protected]: 0.934
+    Task: Body 2D Keypoint
+  Weights: https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/yolox_pose/yoloxpose_l_8xb32-300e_coco-640-de0f8dee_20230829.pth
diff --git a/configs/body_2d_keypoint/yoloxpose/coco/yoloxpose_l_8xb32-300e_coco-640.py b/configs/body_2d_keypoint/yoloxpose/coco/yoloxpose_l_8xb32-300e_coco-640.py
@@ -0,0 +1,17 @@
+_base_ = './yolopose_s_8xb32-300e_coco-640.py'
+
+widen_factor = 1
+deepen_factor = 1
+checkpoint = 'https://download.openmmlab.com/mmdetection/v2.0/yolox/yolox_' \
+    'l_8x8_300e_coco/yolox_l_8x8_300e_coco_20211126_140236-d3bd2b23.pth'
+
+# model settings
+model = dict(
+    backbone=dict(
+        deepen_factor=deepen_factor,
+        widen_factor=widen_factor,
+        init_cfg=dict(checkpoint=checkpoint),
+    ),
+    neck=dict(
+        in_channels=[256, 512, 1024], out_channels=256, num_csp_blocks=3),
+    head=dict(head_module_cfg=dict(widen_factor=widen_factor)))
diff --git a/configs/body_2d_keypoint/yoloxpose/coco/yoloxpose_m_8xb32-300e_coco-640.py b/configs/body_2d_keypoint/yoloxpose/coco/yoloxpose_m_8xb32-300e_coco-640.py
@@ -0,0 +1,16 @@
+_base_ = './yolopose_s_8xb32-300e_coco-640.py'
+
+widen_factor = 0.75
+deepen_factor = 0.67
+checkpoint = 'https://download.openmmlab.com/mmpose/v1/pretrained_models/' \
+             'yolox_m_8x8_300e_coco_20230829.pth'
+
+# model settings
+model = dict(
+    backbone=dict(
+        deepen_factor=deepen_factor,
+        widen_factor=widen_factor,
+        init_cfg=dict(checkpoint=checkpoint),
+    ),
+    neck=dict(in_channels=[192, 384, 768], out_channels=192, num_csp_blocks=2),
+    head=dict(head_module_cfg=dict(widen_factor=widen_factor)))