This is the python implementation of the datalayer in Convolutional Pose Machines release
- Python2.7
- Opencv2.4.13
- Numpy
- IPython
data_transformer.py
is the implenentaiton like the data_transformer.cpp in the Convolutional Pose Machines release code. But something is difference:
- keep the method of data augmentation
- reduce some params
- remove the center map(if you want to detect mutli person, you can add it)
cpm_data.py
shows how to use data_transformer.py in caffe.
util.py
is the Visualization tools, produced by the Convolutional Pose Machines release
- Replace the
data_dir
to your dataset - Refer the data format in example folder to generate your dataset
@inproceedings{wei2016cpm,
author = {Shih-En Wei and Varun Ramakrishna and Takeo Kanade and Yaser Sheikh},
booktitle = {CVPR},
title = {Convolutional pose machines},
year = {2016}
}