Deep Spatial Autoencoders in PyTorch
Original paper: "Deep Spatial Autoencoders for Visuomotor Learning"
Chelsea Finn, Xin Yu Tan, Yan Duan, Trevor Darrell, Sergey Levine, Pieter Abbeel
Available at: https://arxiv.org/pdf/1509.06113.pdf
dsae.py implements the spatial soft-argmax operation, as well as the autoencoder encoder and decoder networks from the original paper. Moreover, we implement the gslow loss contribution as presented in the paper.