Our model architecture is defined as depicted bellow, please refer to the paper for more details:
- python 3.x
- pytorch 4.0+
- Runing the following command to translate edges to shoes&handbags (the pretrained models are stored in ./checkpoints/edges_shoes&handbags directory):
python ./test.py --name edges_shoes&handbags --d_num 2Then the translated samples are stored in ./checkpoints/edges_shoes&handbags/edges_shoes&handbags_results directory. By default, it produce 5 random translation outputs.
- Download the dataset you want to use and move to ./datasets. For example, you can use the horse2zebra dataset provided by CycleGAN. Please make sure that you have the following directory tree structure in your repository
├── datasets
│ └── horse2zebra
│ ├── trainA
│ ├── testA
│ ├── trainB
│ ├── testB
The Animals With Attributes (AWA) dataset can be downloaded from hear.
- Start training with the following command:
python ./train.py --name horse2zebra --d_num 2Intermediate image outputs and model binary files are stored in ./checkpoints/horse2zebra/web
If this work helps to easy your research, please cite this paper :
@article{huang2022multimodal,
title={Multimodal image-to-image translation via a single generative adversarial network},
author={Huang, Shihua and He, Cheng and Cheng, Ran},
journal={IEEE Transactions on Artificial Intelligence},
year={2022},
publisher={IEEE}
}
The code used in this research is based on SingleGAN and CycleGAN.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent