swook / GazeML

Gaze Estimation using Deep Learning, a Tensorflow-based framework.
MIT License
507 stars 141 forks source link

can_use_eye, can_use_eyelid and can_use_iris in elg_demo.py are False #78

Open michealyu opened 3 years ago

michealyu commented 3 years ago

We found that all of these 3 parameters are False so that eye iris, eyeball center and gaze could not be found. Can someone give some suggestions? Thanks a lot!!!!!

keishatsai commented 3 years ago

Because the confidence are not good enough to be beyond in certain limit, you can set all of them are True to show the prediction results. However, the results may not be as good as expected. Firstly, you should make sure that the model has been loaded properly. Secondly, if you done with first one, but still get the same situation. You can lower down the lower limit for eye iris, eyeball and gaze. Lastly, this model meet some tough cases, you can retrain the model or just try another sample.