cchamber/openpose_keras

keras version of Realtime Multi-Person Pose Estimation modified for transfer learning

About this fork

This repo was forked from the Anatolix fork of Realtime Multi-Person Pose Estimation.

For keras version of original Realtime Multi-Person Pose Estimation repository, see the Michal Faber fork and the Anatolix fork

We have edited the Anatolix fork for transfer learning, starting with the trained CMU model weights.

Changes to Anatolix fork

• Add config files to main folder
• Add video demo
• Remove segmentation mask from coco_masks_hdf5.py (replace with bounding box)
• load cmu model weights in train_pose.py

Results

How to run demo/training

• To use COCO: Download the data set (~25 GB) cd dataset; sh get_dataset.sh,

• Or add own data

• Download COCO official toolbox in dataset/coco/ .

• cd coco/PythonAPI; sudo python setup.py install to install pycocotools.

• Use coco-api to view data

• Download converted CMU keras model to model folder

• cd /model;

• sudo ./get_keras_model.sh

Testing steps

Run demo on image

• cd ..
• python3 demo_image.py --image sample_images/ski.jpg
• Output saved in result.png in main folder

Run demo on video

• python3 demo_video.py
• Output saved to video_data folder: video and x,y coordinates of keypoints in pkl file

Training steps

Create .h5 data files

• Edit /training/coco_masks_hdf5.py
  • #!/usr/bin/env python point to python env
  • Point to correct .h5 data files
  • Set size of validation set
• cd training
• Run ./coco_masks_hdf5.py to generate .h5 training files

Run training

• Edit /training/train_pose.py
  • #!/usr/bin/env python point to python env
  • Select gpus
  • Edit batch size if needed
  • Select model file to train on
• Run ./train_pose.py

Model files saved in /training/Canonical

Related repository

• CVPR'16, Convolutional Pose Machines.
• CVPR'17, Realtime Multi-Person Pose Estimation.

Citation

Please cite the paper in your publications if it helps your research:

@InProceedings{cao2017realtime,
  title = {Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields},
  author = {Zhe Cao and Tomas Simon and Shih-En Wei and Yaser Sheikh},
  booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  year = {2017}
  }