Estimating human gaze from natural eye images.
- Python 3.6.8
- PyTorch 1.3.0
Extract images, scan them, to get bounding boxes and landmarks:
$ python extract.py
$ python pre_process.py- Python 3.6.8
- PyTorch 1.3.0
$ python train.pyTo visualize the training process：
$ tensorboard --logdir=runs$ python demo.py| Image | True | Out | Plot |
|---|---|---|---|
 |
look_vec: 0.2209,0.4703,-0.8544 pupil_size: 0.0233 |
look_vec: 0.1917,0.4689,-0.8622 pupil_size: 0.0301 |
 |
 |
look_vec: 0.1296,-0.0196,-0.9914 pupil_size: -0.3122 |
look_vec: 0.2225,0.0327,-0.9744 pupil_size: -0.2559 |
 |
 |
look_vec: -0.6035,0.0833,-0.7930 pupil_size: 0.2096 |
look_vec: -0.6263,0.0704,-0.7764 pupil_size: 0.2214 |
 |
 |
look_vec: -0.5143,-0.2162,-0.8299 pupil_size: 0.0997 |
look_vec: -0.5563,-0.2640,-0.7879 pupil_size: 0.0541 |
 |
 |
look_vec: 0.3604,0.0302,-0.9323 pupil_size: 0.1686 |
look_vec: 0.3937,0.0203,-0.9190 pupil_size: 0.2245 |
 |
 |
look_vec: 0.6021,0.6270,-0.4943 pupil_size: 0.1146 |
look_vec: 0.5902,0.6401,-0.4918 pupil_size: 0.0198 |
 |
 |
look_vec: -0.0691,-0.2175,-0.9736 pupil_size: -0.0675 |
look_vec: -0.0955,-0.1726,-0.9804 pupil_size: -0.0526 |
 |
 |
look_vec: -0.4191,0.3765,-0.8262 pupil_size: -0.3669 |
look_vec: -0.3618,0.3485,-0.8646 pupil_size: -0.2192 |
 |
 |
look_vec: 0.0711,-0.1398,-0.9876 pupil_size: 0.2395 |
look_vec: 0.0902,-0.1491,-0.9847 pupil_size: 0.2621 |
 |
 |
look_vec: 0.3328,-0.4387,-0.8347 pupil_size: -0.2609 |
look_vec: 0.3543,-0.4874,-0.7981 pupil_size: -0.2140 |
 |