---
license: mit
tags:
  - change-detection
  - image-registration
  - optical-flow
  - diffusion-features
  - remote-sensing
library_name: pytorch
pipeline_tag: image-segmentation
---

# DiffRegCD — Pretrained Weights

Pretrained checkpoints for **DiffRegCD: Integrated Registration and Change Detection with
Diffusion Features** (WACV 2026).

- 📄 Paper: https://arxiv.org/abs/2511.07935
- 💻 Code: https://github.com/Anita-Madani/DiffRegCD-Integrated-Registration-and-Change-Detection-with-Diffusion-Features

## Files

| File | Description |
|------|-------------|
| `I200000_E3_gen.pth` | Frozen DDPM backbone (diffusion feature extractor, shared by all datasets) |
| `levir/best_reg_model_gen.pth`, `levir/best_cd_model_gen.pth` | LEVIR-CD registration + change heads |
| `dsifn/…` | DSIFN-CD registration + change heads |
| `whu/…` | WHU-CD registration + change heads |
| `sysu/…` | SYSU-CD registration + change heads |
| `vl_cmu_cd/…` | VL-CMU-CD registration + change heads |

## Usage

```bash
git clone https://github.com/Anita-Madani/DiffRegCD-Integrated-Registration-and-Change-Detection-with-Diffusion-Features.git
cd DiffRegCD-*
pip install -U huggingface_hub
bash scripts/download_weights.sh        # pulls everything here into ./checkpoints/
python main.py -c config/levir_test.json -p test
```

## Citation

```bibtex
@article{madani2025diffregcd,
  title   = {DiffRegCD: Integrated Registration and Change Detection with Diffusion Features},
  author  = {Madani, Seyedehanita and Chellappa, Rama and Patel, Vishal M.},
  journal = {arXiv preprint arXiv:2511.07935},
  year    = {2025}
}
```