Pixel-to-Pixel project

This is the project about Pixel-to-Pixel task. Give an example about cycle-gan and pix2pix. NOT OFFICAL REPOSITORY!!!

This project can also serve as a structure for deep learning projects.

Project Structure

├── data
│   ├── __init__.py
│   ├── image_folder.py
│   ├── base_dataset.py
│   ├── aligned_dataset.py
│   ├── unaligned_dataset.py
├── models
│   ├── __init__.py
│   ├── base_model.py
│   ├── cycle_gan_model.py
│   ├── pix2pix_model.py
├── options
│   ├── __init__.py
│   ├── base_options.py
│   ├── train_options.py
├── util
│   ├── __init__.py
│   ├── logger.py
│   ├── util.py
│   ├── image_pool.py
│   ├── visualizer.py
├── train.py
├── trainer.py

Requirements

• Python 3.11
• PyTorch 2.0.0
• torchvision

Usage

CycleGAN

Image input size: (256, 256)

Download dataset

bash ./datasets/download_cyclegan_dataset.sh horse2zebra

Train

python train.py --dataroot ./datasets/horse2zebra --name horse2zebra --model cycle_gan

Pix2Pix

Image input size: (512, 256)

Download dataset

bash ./datasets/download_pix2pix_dataset.sh facades

Train

python train.py --dataroot ./datasets/facades --name facades_pix2pix --model pix2pix --direction BtoA --batch_size 1024

---

Acknowledgments

This project is inspired by cycle-gan. Thanks for the author's contribution.