SimGAN_PyTorch

A Simple Pytorch Implementation of Apple's Learning from Simulated and Unsupervised Images through Adversarial Training

This code is based on AlexHex7/SimGAN_pytorch(PyTorch) and inspired by wayaai/SimGAN(Keras&Tensorflow).

I try to re-organize AlexHex7's code, fix some bugs and speed up the training procedure.

Time Cost

With [Batch Size=128, k_d=1k_g=50], it takes about 12 hours to finish 10000 training steps.

Current Results

Note that my parameters and optimizers are different from the original work.

Results after 10000 iteration (About 1/4 of total steps in the original work)