This is a Matlab/Mex implementation of a 2D human pose estimator that was used in the paper "Multi-view Body Part Recognition with Random Forests".
For more information please refer to my homepage: http://www.csc.kth.se/~vahidk/
If you plan to use this source code please cite:
@inproceedings{kazemi2013multi,
title={Multi-view Body Part Recognition with Random Forests},
author={Kazemi, Vahid and Burenius, Magnus and Azizpour, Hossein and Sullivan, Josephine},
booktitle={BMVC},
year={2013}
}
You need matlab to run this code. To compile and run the code simply run:
cd {RFPoseEstimationDirectory}/code
make
globals
football_test
```
This runs the code for testing the model on the pretrained football dataset.
To train a new model run:
```
globals
football_train
```
You can change the code to read your own dataset. Look at how it's done in
football_train.m and football_test.m
Vahid Kazemi