PPAP

By Jiyong Rao, Brian Nlong Zhao and Yu Wang

This repository is an official implementation of PPAP in the paper PPAP: Probabilistic Prompt Distribution Learning for Animal Pose Estimation, which is accepted to CVPR 2025.

Installation

Requirements

• Linux, CUDA>=9.2, GCC>=5.4

• Python>=3.6

• PyTorch>=1.5.0, torchvision>=0.6.0 (following instructions here)

• mmcv

  cd mmcv
  pip install -r requirements.txt
  pip install -v -e .
  

• mmpose

  cd ..
  pip install -r requirements.txt
  pip install -v -e .
  

Usage

Dataset preparation

Please download the dataset from AP-10K.

CLIP-pretrained models

Please download CLIP pretrained models from CLIP.

Training

Training PPAP on AP-10K

bash run_GCAP_ViTB.sh

Acknowledgement

This project is based on mmpose, AP-10K, CLIP, and DenseCLIP. Thanks for their wonderful works. See LICENSE for more details.

Citing PPAP

If you find PPAP useful in your research, please consider citing:

@inproceedings{rao2025probabilistic,
  title={Probabilistic Prompt Distribution Learning for Animal Pose Estimation},
  author={Rao, Jiyong and Zhao, Brian Nlong and Wang, Yu},
  booktitle=CVPR,
  pages={29438--29447},
  year={2025}
}