jerrywiston / lightweight-human-pose-estimation.pytorch

Fast and accurate human pose estimation in PyTorch. Contains implementation of "Real-time 2D Multi-Person Pose Estimation on CPU: Lightweight OpenPose" paper.

Openpose ADDA

Table of Contents

• Prerequisites
• Dataset
• Training
• Validation
• Pre-trained model
• Python demo
• Citation

Prerequisites

2. Install requirements pip install -r requirements.txt

Dataset

• ADDA Dataset Link
  • RGB
  • Thermal

Training

Train ADDA

python3 train_adda.py

• Any configuration about training in config.py
  • Include source folder path and target folder path

Validation

1. Run python val.py --labels <COCO_HOME>/annotations/person_keypoints_val2017.json --images-folder <COCO_HOME>/val2017 --checkpoint-path <CHECKPOINT>

Pre-trained model

The model expects normalized image (mean=[128, 128, 128], scale=[1/256, 1/256, 1/256]) in planar BGR format. Pre-trained on COCO model is available at: https://download.01.org/opencv/openvino_training_extensions/models/human_pose_estimation/checkpoint_iter_370000.pth, it has 40% of AP on COCO validation set (38.6% of AP on the val subset).

Python Demo

We provide python demo just for the quick results preview. Please, consider c++ demo for the best performance. To run the python demo from a webcam:

• python demo.py --checkpoint-path <path_to>/checkpoint_iter_370000.pth --video 0

Citation:

If this helps your research, please cite the paper:

@inproceedings{osokin2018lightweight_openpose,
    author={Osokin, Daniil},
    title={Real-time 2D Multi-Person Pose Estimation on CPU: Lightweight OpenPose},
    booktitle = {arXiv preprint arXiv:1811.12004},
    year = {2018}
}