Copyright (C) 2018 NVIDIA Corporation. All rights reserved. Licensed under the CC BY-NC-SA 4.0 license (https://creativecommons.org/licenses/by-nc-sa/4.0/legalcode).
pytorch, yaml, tensorboard (from https://github.com/dmlc/tensorboard), and tensorboardX (from https://github.com/lanpa/tensorboard-pytorch).
The code base was developed using Anaconda with the following packages.
conda install pytorch=0.3 torchvision cuda80 -c pytorch;
conda install -y -c anaconda pip;
conda install -y -c anaconda pyyaml;
pip install tensorboard tensorboardX;
We also provide a Dockerfile for building an environment for running the MUNIT code.
First, download the pretrained models and put them in models folder.
Run the following command to translate edges to shoes
python test.py --config configs/edges2shoes_folder.yaml --input inputs/edge.jpg --output_folder outputs --checkpoint models/edges2shoes.pt --a2b 1
The results are stored in outputs folder. By default, it produces 10 random translation outputs.
The above command outputs diverse shoes from an edge input. In addition, it is possible to control the style of output using an example shoe image.
python test.py --config configs/edges2shoes_folder.yaml --input inputs/edge.jpg --output_folder outputs --checkpoint models/edges2shoes.pt --a2b 1 --style inputs/shoe.jpg
-
Download the dataset you want to use. For example, you can use the edges2shoes dataset provided by Zhu et al.
-
Setup the yaml file. Check out
configs/edges2handbags_folder.yamlfor folder-based dataset organization. Change thedata_rootfield to the path of your downloaded dataset. For list-based dataset organization, check outconfigs/edges2handbags_list.yaml -
Start training
python train.py --config configs/edges2handbags_folder.yaml -
Intermediate image outputs and model binary files are stored in
outputs/edges2handbags_folder