A deep learning model to detect state of eye (open/close) in image/video
The model was trained on this dataset of closed and open eyes.
To predict for a video use function predict_on_video and to predict for an image use predict_on_image in the Jupyter notebook.
Below is a sample prediction for the "Coca-Cola | Eyes Closed" video by Coca-Cola.
The prediction of the model (open or close) is shown in green in the top-left of video, with the confidence in parenthesis.
The model is trained only to detect open and closed eyes. Hence it gives a closed eye prediction when the image/video frame does not have a face.