TensorFlow 1.x Deep Learning Cookbook

Creating a DeepDream network

Google trained a neural network on ImageNet for the Large Scale Visual Recognition Challenge (ILSVRC) in 2014 and made it open source in July 2015. The original algorithm is presented in Going Deeper with Convolutions, Christian Szegedy, Wei Liu, Yangqing Jia, Pierre Sermanet, Scott Reed, Dragomir Anguelov, Dumitru Erhan, Vincent Vanhoucke e Andrew Rabinovich (https://arxiv.org/abs/1409.4842) . The network learned a representation of each image. The lower layers learned low-level features, such as lines and edges, while the higher layers learned more sophisticated patterns such as eyes, noses, mouths, and so on. Therefore, if we try to represent a higher level in the network, we will see a mix of different features ...