EvaNet [TPAMI 2026]

This is the offical implementation for the paper titled "EvaNet: Towards More Efficient and Consistent Infrared and Visible Image Fusion Assessment"(Arxiv, Paper).

---

Announcement

• Apr 03, 2026: Our paper has been accepted by IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI).
• Apr 14, 2026: The online evaluation platform EvaJudge is now available: 👉 http://evanet.online:5001
• Apr 21, 2026: Results of TDFusion (CVPR 2025) have been added to the Public Leaderboard.
• Apr 21, 2026: (v1.1.3) Support visualization: click a method name to view its fusion examples.
• May 11, 2026: Results of UPTP (CVPR 2026) have been added to the Public Leaderboard.
• May 12, 2026: Results of FusionRegister (CVPR 2026) have been added to the Public Leaderboard.
• May 15, 2026: Results of TEDFusion (ICML 2026) have been added to the Public Leaderboard.
• May 16, 2026: Results of SMLNet (IJCV 2025, first attempt of manifold learning) have been added to the Public Leaderboard.
• May 27, 2026: Results of ISFL (cvpr 2026, an intervention-based method) have been added to the Public Leaderboard.
• Jun 05, 2026: Results of LTOFusion (TIP 2026, a Learning-To-Optimize framework with flow matching) have been added to the Public Leaderboard.

---

EvaJudge Usage

EvaJudge is an online evaluation platform for infrared and visible image fusion, powered by EvaNet. It enables fast, consistent, and zero-setup evaluation of fusion results.

You are welcome to submit your papers, results, and model weights for inclusion in the Public Leaderboard..

Benchmark datasets can be founded here.

📮Email: chunyang_cheng@jiangnan.edu.cn ☎️WeChat: chengchunyang2016

🔹 How to Use

1. Register an account here.
2. Log in and select a dataset (e.g., LLVIP, MSRS).
3. Upload your fusion results as a .zip file.
4. The system will automatically evaluate your results and return a full set of metrics within seconds.

🔹 Submission Format

• The uploaded file must be a .zip archive.
• The archive should contain only fused images (no subfolders).
• File names must correspond to the dataset image indices (e.g., 1.jpg, 2.jpg, ..., N.jpg).

🔹 Registration Policy

• Registration is restricted to institutional (organization-affiliated) email addresses only (e.g., university or company domains).

Citation

If this work is helpful to you, please cite it as:

@article{ChengEvaNetTPAMI2026,
  author={Cheng, Chunyang and Xu, Tianyang and Wu, Xiao-Jun and Zhou, Tao and Li, Hui and Tang, Zhangyong and Kittler, Josef},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence}, 
  title={EvaNet: towards More Efficient and Consistent Infrared and Visible Image Fusion Assessment}, 
  year={2026},
  volume={},
  number={},
  pages={1-18},
  keywords={Image fusion;quality assessment;efficient;unified;large language model},
  doi={10.1109/TPAMI.2026.3681958}}