Deep networks for group discovery and multi level activity recognition

Requirements: Octave / MATLAB Python 3.5 You'll need only a subset of packages mentioned in requirements.txt (Keras 2.1.4, Tensorflow 1.0.0)

Steps:

1. Run anno_to_csv.m
2. Download dataset images from http://www.eecs.umich.edu/vision/eccv12data/eccv12_images.tar.gz
3. Extract the images folder in the above compressed file and put it into data folder
4. Run pairwise_distance.py to perform training and visualize results.