Convolutional neural network that does real-time emotion recognition. HappyNet detects faces in video and images, classifies the emotion on each face, then replaces each face with the correct emoji for that emotion. Based on Caffe and the "Emotions in the Wild" network available on Caffe model zoo.
I follow the steps you provide strictly, but the loss is -nan from beginning. The training data is properly converted, what is the reason?