Tensorflow 2 and Keras Implementation of EyeTracking for Everyone paper by Torralba et al.
Model was trained on the small dataset of 48K training samples and 5000 validation samples found here with the following configuration:
- Python 3.8
- Tensorflow 2.5.0
- Keras 2.4.3
- Numpy 1.19.5
Results are obtained on 64x64 images. The results are expressed in terms of Mean Absolute Error (MAE) and Standard Deviation. The network was tested on the aforementioned dataset.
| Input Size | MAE | STD | MSE Loss |
|---|---|---|---|
| 64x64 | 0.89, 0.99 | 1.09, 1.22 | 2.2281 |
This work wouldn't be accomplished without the contribution of hugochan and gdubrg