Instruction.txt

To test this project in following steps:
1.Run the demo.m in /pose. And get the human pose estimation image in /database/poseresult/.
you can skip this step ,I have output a demo in /database/poseresult/.

2.When first run this project, you need to train unary model.Run:
	python preprocessing.py -p    (-p means use pretrain model)
This will take some time.

3.Run the mrf,Run:
	python mrf.py
you would see the mrfresult in ./database/mrfresult/



To Run the whole project from scratch:
1.Download all the dataset.
In project dictionary ,run:
	chmod +x getdata.sh
	./getdata.sh
This will download data from dataset and combine into database dictionary

2.Run the demo.m in /pose. And get the human pose estimation image in /database/poseresult/.


3.processing all data,Run: 
	python preprocessing.py

4.Run the mrf,Run:
	python mrf.py
you would see the mrfresult in ./database/mrfresult/



other information.
the pygco need Compile if you download the source file from github.
This is python wrapper for gco-3.0 package. you can follow the instruction of github: https://github.com/yujiali/pygco