Computer Vision News - September 2018

that residual learning aims at predicting gaussian noise and batch normalization helps us normalize the activation function’s output exactly to this range. With highly impressive qualitative results, as you can see below: Research 9 Research Computer Vision News