GramGAN: Deep 3D Texture Synthesis From 2D Exemplars

This repository is the official implementation of GramGAN: Deep 3D Texture Synthesis From 2D Exemplars.

Requirements

• python (2.7/3.7 tested)
• tensorflow
• python-opencv
• scipy

Training

To train the model, run this command:

python train_single.py --training_exemplar <path_to_exemplar>

Evaluation

To evaluate the model, run:

python inference_single.py --model_name <path_to_chpt> --out_path <path_to_output_directory>