UAV-GESTURE: A Dataset for UAV Control and Gesture Recognition
This is a slightly modified version of the original P-CNN implementation (https://github.com/gcheron/P-CNN).
SVM Classification files are from https://github.com/ZhigangTU/HR-MSCNN.
How to run the code:
-
Follow the original installation instructions availabe on https://github.com/gcheron/P-CNN.
-
Download the RGB and FLOW networks from https://www.di.ens.fr/willow/research/p-cnn/download/models.tar, and save them in the "models" folder.
-
Extract images using "video_to_frames.m" file.
-
Run "demo.m". Update "param.impath" variable with extracted images path.
-
Estimated body joints (using OpenPose) are already saved in 'UAVGESTURE/joint_positions'.
-
Please send an email to asanka.perera@mymail.unisa.edu.au to get the video dataset download link.
-
Project website: https://asankagp.github.io/uavgesture/.