martes, 15 de marzo de 2016

DeepDream through all the layers of GoogleNet

Video depicting all the layers of a deep nerual network. The Convolutional Neural Network has been pretrained with ImageNet. The first input image is a picture of myself and at every step the image is zoomed with a ratio of 0.05. At every step the actual input image (frame) is forwarded to the actual hidden layer. The error is set to be the same representation in order to maximize all its activations. Then, a backward pass is computed to modify the input image. After 100 iterations of zooming in one layer, the next layer is used. See more in my Aalto personal web-page 

No hay comentarios:

Publicar un comentario