Trained a CycleGAN Model for Real ↔ Painting Image Style Transfer Using PyTorch

I built and trained a custom CycleGAN model for real ↔ painting style transfer using PyTorch.

It transforms real-world photos into painting-style images — and back — using unpaired datasets.

---

🛠️ Key Features

• ResNet-based Generator + PatchGAN Discriminator
• Cycle-consistency, identity, and adversarial losses
• Full training, validation, and testing pipeline
• Inference on single images or full directories

---

🔗 GitHub Repo

👉 github.com/hitendraa/cyclegan

Includes full code, setup instructions, and example outputs.

---

📖 Based On

Original paper: Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks

Reference repo: junyanz/pytorch-CycleGAN-and-pix2pix

---

Let me know what you think or if you want to try extending it!