CNN(convolutional neural network) was trained on subset of IMDB-WIKI dataset.
CNN has 3 conovolution layer and 2-hidden Layer.
Tracking.js is used for localization of face and then localized facial image is
feeded into CNN.
CNN Architecture Photo
In Case, You're wondering why some of filters are not working. They're called
Dead filters. Dead filters are result of high learning rate.
First Convolution Layer's Activation-Map
Second Convolution Layer's Activation-Map
Third Convolution Layer's Activation-Map