i made a dcnn using keras.
It takes input shape of image 96x96x1. It has 3 conv layers each has POOL-2 after it. It has three FC layers and at the end softmax is used
