Pedestrian Pose Annotation for DukeMTMC-reID by machine (Python and Matlab API)
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
result 0.2 Jan 29, 2018
LICENSE_DukeMTMC-reID.txt Update LICENSE_DukeMTMC-reID.txt Jun 3, 2018
LICENSE_DukeMTMC.txt Update LICENSE_DukeMTMC.txt Jun 3, 2018
README.md Update README.md Feb 1, 2018
demo.png add demo image Jan 29, 2018
generate.py add generation Feb 1, 2018
matlab_demo.m 0.2 Jan 29, 2018
python_demo.ipynb 0.1 Jan 29, 2018

README.md

DukeMTMC-Pose

Pedestrian Pose Annotation for DukeMTMC-reID by machine(Python and Matlab API).

It is a small work but we think it may help the community to focus on the body structure of the pedestrian. The application can be pedestrain hallucination (generation), person re-identification and tracking.

Data

18 pedestrian body points have been included in result folder.

File Description
gallery_points.json The points of gallery images.
train_points.json The points of training images.
query_points.json The points of query images.

We generated the pose by OpenPose. Thank their great works.

(Note that value is null if the point has not been detected.)

API

python: python_demo.ipynb

matlab: matlab_demo.m

Generate the pose by yourself

You need to install the openpose from https://github.com/ZheC/Realtime_Multi-Person_Pose_Estimation. The code is wrote with pycaffe.

I wrote a simple code generate.py to generate the pose. You may directly change the dataset path in it and have a try.

License

Please refer to LICENSE_DukeMTMC.txt and LICENSE_DukeMTMC-reID.txt.