Implementation of the GauGAN Nvidia research paper in Tensorflow rather than Pytorch.
Note: This repository will not contain the ADE20k dataset due to its size. The dataset (ADE20K_2016_07_26/) should be in the data/ directory.
The original ADE20K_2016_07_26/ dataset is prepared for the network by running get_landscape_img.py, which selects images and corresponding segmentation maps from the dataset as follows:
- explicit_cv_landscapes_final_project.txt lists explicit scene categories to include
- objects_we_want.txt lists words and phrases (each phrase on a separate line) that describe objects, and all images that are known to contain at least one of the listed objects are included
Changes Made:
- Shrank image sizes to 128x96
- Reduced the number of upsampling layers in the generator from 7 to 5
- Reduced z-dimension to 64
- We specifically do not include the encoder in our generator. It is in the code, but not used.