🔥 SmokeBench: A Real-World Dataset for Surveillance Image Desmoking in Early-Stage Fire Scenes [ACM MM 2025 Datasets Track]

Wenzhuo Jin* ¹, Qianfeng Yang* ²,Xianhao Wu* ³, Hongming Chen ⁴, Pengpeng Li ⁵, Xiang Chen ⁵

Beijing Jiaotong University¹, Dalian Polytechnic University², Nanjing University of Aeronautics and Astronautics³, Dalian Martime University⁴, Nanjing University of Science and Technology⁵

[Paper]

👉️ Welcome to visit our website (专注底层视觉领域的信息服务平台) for low-level vision:https://lowlevelcv.com/

---

⛳️ To do

• ✔ Release the dataset
• ✔ Release the visual results

---

🔨 Dataset pipeline

---

🔎 Illustration of the SmokeBench

---

⬇️ Dataset Download

Download Link	Description
Google Drive / Baidu Netdisk (refs)	Tran: 9,875 pairs. Test: 100 pairs.

---

📘 Quantitative Results

---

📷️ Visual Results

We provide evaluation code (Python Code) to assess PSNR, SSIM, and LPIPS results. You need to set the parameters --generated_images_path, --target_path, and --Score_save_path.

If WeatherBench is helpful for you, please help star the GitHub Repo. Thanks!

Method	Download Link
FFA-Net	Google Drive / Baidu Netdisk (6666)
MPRNet	Google Drive / Baidu Netdisk (6666)
Restormer	Google Drive / Baidu Netdisk (6666)
Uformer	Google Drive / Baidu Netdisk (6666)
DehazeFormer	Google Drive / Baidu Netdisk (6666)
MB-TaylorFormer	Google Drive / Baidu Netdisk (6666)
X-Restormer	Google Drive / Baidu Netdisk (6666)
MambaIR	Google Drive / Baidu Netdisk (6666)

---

❣ Citation

@article{jin2025smokebench,
  title={SmokeBench: A Real-World Dataset for Surveillance Image Desmoking in Early-Stage Fire Scenes},
  author={Jin, Wenzhuo and Yang, Qianfeng and Wu, Xianhao and Chen, Hongming and Li, Pengpeng and Chen, Xiang},
  journal={arXiv preprint arXiv:2509.12701},
  year={2025}
}

---

📧 Contact

If you have any questions, please feel free to contact [email protected] or [email protected].