Reimplementation in python of matlab code from http://gazefollow.csail.mit.edu/
Download model weights (.mat, .prototxt and .caffemodel files) from http://gazefollow.csail.mit.edu/download.html and put them into ./utils/
Dependencies (tested with Python 3.6 but later versions should work):
- Numpy
- Opencv 4
- Scipy (for reading matlab files within python
Visualize GazeFollow dataset
python test_gazefollow_reader.py <path/to/gazefollow/train_annotations.txt>
Test the network on a custom picture, manually giving eye position (x and y in (0,1))
python test_gazefollow_network.py <path/to/picture.png> <eye_x> <eye_y>
Test the network on a random pictures from the gazefollow dataset
python test_gazefollow_network_2.py <path/to/gazefollow/train_annotations.txt>