E3Diff: SAR-to-Optical Image Translation
🏆 1st Place - CVPR PBVS2025 Multi-modal Aerial View Image Challenge
Model Description
E3Diff is an efficient end-to-end diffusion model for one-step SAR-to-Optical translation.
Key Features
- Real-time inference: 0.17s per 256x256 image on A6000
- High quality: 35% FID improvement over previous SOTA
- One-step sampling: Unlike traditional diffusion (1000 steps)
Usage
from huggingface_hub import hf_hub_download
import torch
# Download weights
weights = hf_hub_download(repo_id="Dhenenjay/E3Diff-SAR2Optical", filename="I700000_E719_gen.pth")
Citation
@ARTICLE{10767752,
author={Qin, Jiang and Zou, Bin and Li, Haolin and Zhang, Lamei},
journal={IEEE Geoscience and Remote Sensing Letters},
title={Efficient End-to-End Diffusion Model for One-step SAR-to-Optical Translation},
year={2024},
doi={10.1109/LGRS.2024.3506566}
}