Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[Doc] Refine SimCC and IPR algorithm readme with model performance table #1710

Merged
merged 2 commits into from
Oct 10, 2022
Merged

[Doc] Refine SimCC and IPR algorithm readme with model performance table #1710

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
6e043eb
add algorithm readme
Tau-J Oct 10, 2022
75a215f
update ckpt
Tau-J Oct 10, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 15 additions & 0 deletions configs/body_2d_keypoint/integral_regression/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,15 @@
# Top-down integral-regression-based pose estimation

Top-down methods divide the task into two stages: object detection, followed by single-object pose estimation given object bounding boxes. At the 2nd stage, integral regression based methods use a simple integral operation relates and unifies the heatmap and joint regression differentiably, thus obtain the keypoint coordinates given the features extracted from the bounding box area, following the paradigm introduced in [Integral Human Pose Regression](https://arxiv.org/abs/1711.08229).

## Results and Models

### COCO Dataset

Results on COCO val2017 with detector having human AP of 56.4 on COCO val2017 dataset

| Model | Input Size | AP | AR | Details and Download |
| :------------------: | :--------: | :---: | :---: | :---------------------------------------------------: |
| ResNet-50+Debias-IPR | 256x256 | 0.675 | 0.765 | [resnet_debias_coco.md](./coco/resnet_debias_coco.md) |
| ResNet-50+DSNT | 256x256 | 0.674 | 0.764 | [resnet_dsnt_coco.md](./coco/resnet_dsnt_coco.md) |
| ResNet-50+IPR | 256x256 | 0.633 | 0.730 | [resnet_ipr_coco.md](./coco/resnet_ipr_coco.md) |
20 changes: 20 additions & 0 deletions configs/body_2d_keypoint/simcc/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,20 @@
# Top-down SimCC-based pose estimation

Top-down methods divide the task into two stages: object detection, followed by single-object pose estimation given object bounding boxes. At the 2nd stage, SimCC based methods reformulate human pose estimation as two classification tasks for horizontal and vertical coordinates, and uniformly divide each pixel into several bins, thus obtain the keypoint coordinates given the features extracted from the bounding box area, following the paradigm introduced in [SimCC: a Simple Coordinate Classification Perspective for Human Pose Estimation](https://arxiv.org/abs/2107.03332).

<div align=center>
<img src="https://user-images.githubusercontent.com/13503330/189811385-6395d118-055b-4bad-89e8-f84ffa2c2aa6.png">
</div>

## Results and Models

### COCO Dataset

Results on COCO val2017 with detector having human AP of 56.4 on COCO val2017 dataset

| Model | Input Size | AP | AR | Details and Download |
| :---------------------------: | :--------: | :---: | :---: | :-----------------------------------------------: |
| ResNet-50+SimCC | 384x288 | 0.735 | 0.790 | [resnet_coco.md](./coco/resnet_coco.md) |
| ResNet-50+SimCC | 256x192 | 0.721 | 0.781 | [resnet_coco.md](./coco/resnet_coco.md) |
| S-ViPNAS-MobileNet-V3+SimCC | 256x192 | 0.695 | 0.755 | [vipnas_coco.md](./coco/vipnas_coco.md) |
| MobileNet-V2+SimCC(wo/deconv) | 256x192 | 0.620 | 0.678 | [mobilenetv2_coco.md](./coco/mobilenetv2_coco.md) |
2 changes: 1 addition & 1 deletion configs/body_2d_keypoint/simcc/coco/mobilenetv2_coco.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -52,4 +52,4 @@ Results on COCO val2017 with detector having human AP of 56.4 on COCO val2017 da

| Arch | Input Size | AP | AP<sup>50</sup> | AP<sup>75</sup> | AR | AR<sup>50</sup> | ckpt | log |
| :-------------------------------------------- | :--------: | :---: | :-------------: | :-------------: | :---: | :-------------: | :-------------------------------------------: | :-------------------------------------------: |
| [simcc_mobilenetv2_wo_deconv](/configs/body_2d_keypoint/simcc/coco/simcc_mobilenetv2_wo-deconv-8xb64-210e_coco-256x192.py) | 256x192 | 0.620 | 0.855 | 0.697 | 0.678 | 0.902 | [ckpt](https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/simcc/coco/simcc_mobilenetv2_wo-deconv-8xb64-210e_coco-256x192-e0cc028d_20220922.pth) | [log](https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/simcc/coco/simcc_mobilenetv2_wo-deconv-8xb64-210e_coco-256x192-e0cc028d_20220922.log.json) |
| [simcc_mobilenetv2_wo_deconv](/configs/body_2d_keypoint/simcc/coco/simcc_mobilenetv2_wo-deconv-8xb64-210e_coco-256x192.py) | 256x192 | 0.620 | 0.855 | 0.697 | 0.678 | 0.902 | [ckpt](https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/simcc/coco/simcc_mobilenetv2_wo-deconv-8xb64-210e_coco-256x192-4b0703bb_20221010.pth) | [log](https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/simcc/coco/simcc_mobilenetv2_wo-deconv-8xb64-210e_coco-256x192-4b0703bb_20221010.log.json) |
2 changes: 1 addition & 1 deletion docs/src/papers/algorithms/simcc.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
# Simple Coordinate Classification Perspective for Human Pose Estimation
# SimCC: a Simple Coordinate Classification Perspective for Human Pose Estimation

<!-- [ALGORITHM] -->

Expand Down