CoDE-GAN: Content Decoupled and Enhanced GAN for Sketch-guided Flexible Fashion Editing

codegan-demo.mp4

This is the official implementation of CoDE-GAN (ToMM 2025). The trained weights and a more detailed README are in progress.

Environment Requirements

We have tested the requirements on torch==2.4.1 in cuda-11.8. But we believe this repository should have no speciall requirements for torch or cuda versions.

1. Create a conda environment and install a torch that compatible with your cuda.

conda create -n codegan python=3.10
pip install torch==2.4.1 torchvision==0.19.1 torchaudio==2.4.1 --index-url https://download.pytorch.org/whl/cu118

2. Install mmcv for dependencies.

pip install openmim
mim install mmcv-full

3. Install other requirements.

pip install -r requirements.txt

Inference

python inference-img.py