Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,48 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
tags:
|
| 4 |
+
- image-to-image
|
| 5 |
+
- SAR
|
| 6 |
+
- remote-sensing
|
| 7 |
+
- diffusion
|
| 8 |
+
- computer-vision
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
# E3Diff: SAR-to-Optical Image Translation
|
| 12 |
+
|
| 13 |
+
**🏆 1st Place - CVPR PBVS2025 Multi-modal Aerial View Image Challenge**
|
| 14 |
+
|
| 15 |
+
## Model Description
|
| 16 |
+
|
| 17 |
+
E3Diff is an efficient end-to-end diffusion model for one-step SAR-to-Optical translation.
|
| 18 |
+
|
| 19 |
+
### Key Features
|
| 20 |
+
- **Real-time inference**: 0.17s per 256x256 image on A6000
|
| 21 |
+
- **High quality**: 35% FID improvement over previous SOTA
|
| 22 |
+
- **One-step sampling**: Unlike traditional diffusion (1000 steps)
|
| 23 |
+
|
| 24 |
+
## Usage
|
| 25 |
+
|
| 26 |
+
```python
|
| 27 |
+
from huggingface_hub import hf_hub_download
|
| 28 |
+
import torch
|
| 29 |
+
|
| 30 |
+
# Download weights
|
| 31 |
+
weights = hf_hub_download(repo_id="Dhenenjay/E3Diff-SAR2Optical", filename="I700000_E719_gen.pth")
|
| 32 |
+
```
|
| 33 |
+
|
| 34 |
+
## Citation
|
| 35 |
+
|
| 36 |
+
```bibtex
|
| 37 |
+
@ARTICLE{10767752,
|
| 38 |
+
author={Qin, Jiang and Zou, Bin and Li, Haolin and Zhang, Lamei},
|
| 39 |
+
journal={IEEE Geoscience and Remote Sensing Letters},
|
| 40 |
+
title={Efficient End-to-End Diffusion Model for One-step SAR-to-Optical Translation},
|
| 41 |
+
year={2024},
|
| 42 |
+
doi={10.1109/LGRS.2024.3506566}
|
| 43 |
+
}
|
| 44 |
+
```
|
| 45 |
+
|
| 46 |
+
## Links
|
| 47 |
+
- [Paper](https://ieeexplore.ieee.org/document/10767752)
|
| 48 |
+
- [GitHub](https://github.com/DeepSARRS/E3Diff)
|