DistilPose: Tokenized Pose Regression with Heatmap Distillation

Implementation for: DistilPose: Tokenized Pose Regression with Heatmap Distillation

DistilPose: Tokenized Pose Regression with Heatmap Distillation,
Suhang Ye*, Yingyi Zhang*, Jie Hu*, Liujuan Cao, Shengchuan Zhang^✉, Lei Shen, Jun Wang, Shouhong Ding, Rongrong Ji.
In: Conference on Computer Vision and Pattern Recognition~(CVPR), 2023
arXiv preprint (arXiv 2303.02455)
(* equal contribution)

Introdoction

In the field of human pose estimation, regression-based methods have been dominated in terms of speed, while heatmap-based methods are far ahead in terms of performance. How to take advantage of both schemes remains a challenging problem. In this paper, we propose a novel human pose estimation framework termed DistilPose, which bridges the gaps between heatmap-based and regression-based methods.

Our contributions are summarized as follows:

• We propose a novel human pose estimation framework, DistilPose, which is the first work to transfer knowledge between heatmap-based and regression-based models losslessly.
• We introduce a novel Token-distilling Encoder (TDE) to take advantage of both heatmap-based and regression-based models. With the proposed TDE, the gap between the output space of heatmaps and coordinate vectors can be facilitated in a tokenized manner.
• We propose Simulated Heatmaps to model explicit heatmap information, including 2D keypoint distributions and keypoint confidences. With the aid of Simulated Heatmaps, we can transform the regression-based HPE task into a more straightforward learning task that fully exploits local information. Simulated Heatmaps can be applied to any heatmap-based and regression-based models for transferring heatmap knowledge to regression models.

Installation & Quick Start

DistilPose depends on PyTorch and MMPose, please install following packages:

conda create -n distilpose python=3.8 pytorch=1.7.0 torchvision -c pytorch -y
conda activate distilpose
pip3 install openmim
mim install mmcv-full==1.3.8
git submodule update --init
cd mmpose
git checkout v0.22.0
pip3 install -e .
cd ..
pip3 install -r requirements.txt

Download teacher checkpoint and save to "./teacher_chkpts". Train model on COCO as below:

./tools/dist_train.sh configs/body/2d_kpt_sview_rgb_img/distilpose/coco/DistilPose_S_coco_256x192.py 8

For evaluating on COCO, downlowd checkpoint and run the following command lines:

./tools/dist_test.sh configs/body/2d_kpt_sview_rgb_img/distilpose/coco/DistilPose_S_coco_256x192.py \
./checkpoints/distilpose_s.pth 8

Models

COCO Keypoint Detection Results

Name	Role	Param	GFLOPs	AP	download
DistilPose-S	Student	5.4	2.38	71.6	Code:3733
DistilPose-L	Student	21.3	10.33	74.4	Code:5tni
TokenPose-L	Teacher	69.4	17.03	75.2	Code:b8vn

Citations

Please consider citing our papers in your publications if the project helps your research. BibTeX reference is as follows.

@article{ye2023distilpose,
  title={DistilPose: Tokenized Pose Regression with Heatmap Distillation},
  author={Ye, Suhang and Zhang, Yingyi and Hu, Jie and Cao, Liujuan and Zhang, Shengchuan and Shen, Lei and Wang, Jun and Ding, Shouhong and Ji, Rongrong},
  journal={arXiv preprint arXiv:2303.02455},
  year={2023}
}