Pytorch implementation of DeePSiM

This repository contains a Pytorch-based implementation of DeePSiM [Dosovitskiy & Brox, 2016]. The code here is desigined for inverting a VGG16-like network trained on CIFAR10 dataset, but you can modify slightely to work for other network architectures and datasets.

Tested environment

• Python 3.6.6
• Pytorch 1.0.1

Usage

1. run ff_training.py to train the feed-forward network on CIFAR10 (input --> target)
2. run get_rpr.py to obtain the internal representation at the specified layer (input --> hidden).
3. run invert.py to invert the internal representation back to the input space (hidden --> input).

Example results

Reference

Generating Images with Perceptual Similarity Metrics based on Deep Networks https://arxiv.org/abs/1602.02644.

Hyperparameter values are obtained with reference to the above paper and the Tensorflow implementation.