Rethinking Surgical Smoke:
A Type-Aware Laparoscopic Video Desmoking Method and Dataset (Dataset Website)

Accepted at AAAI 2026

Qifan Liang1,2,5, Junlin Li3, Zhen Han*1,2, Xihao Wang1,2, Zhongyuan Wang1,2, Bin Mei4

1 National Engineering Research Center for Multimedia Software, School of Computer Science, Wuhan University, China

2 Hubei Key Laboratory of Multimedia and Network Communication Engineering, School of Computer Science, Wuhan University, China

3 School of Cyber Science and Engineering, Wuhan University, China

4 Zhongnan Hospital, Wuhan University, China

5 National University of Singapore, Singapore

* Corresponding author

Paper Website Baidu Netdisk (Password: AAAI) Google Drive Arxiv Paper

Visual Demo

Desmoking Demo

Dataset Overview: From top to bottom are Diffusion Smoke, Ambient Smoke, and their entanglement, along with their corresponding GT smoke masks.

Abstract

Electrocautery or lasers will inevitably generate surgical smoke, which hinders the visual guidance of laparoscopic videos for surgical procedures. The surgical smoke can be classified into different types based on its motion patterns, leading to distinctive spatio-temporal characteristics across smoky laparoscopic videos. However, existing desmoking methods fail to account for such smoke-type-specific distinctions. Therefore, we propose the first Smoke-Type-Aware Laparoscopic Video Desmoking Network (STANet) by introducing two smoke types: Diffusion Smoke and Ambient Smoke. Specifically, a smoke mask segmentation sub-network is designed to jointly conduct smoke mask and smoke type predictions based on the attention-weighted mask aggregation, while a smokeless video reconstruction sub-network is proposed to perform specially desmoking on smoky features guided by two types of smoke mask. To address the entanglement challenges of two smoke types, we further embed a coarse-to-fine disentanglement module into the mask segmentation sub-network, which yields more accurate disentangled masks through the smoke-type-aware cross attention between non-entangled and entangled regions. In addition, we also construct the first large-scale synthetic video desmoking dataset with smoke type annotations. Extensive experiments demonstrate that our method not only outperforms state-of-the-art approaches in quality evaluations, but also exhibits superior generalization across multiple downstream surgical tasks.

Dataset Structure

The dataset is organized hierarchically. Below is the directory tree structure for STSVD.

Directory Tree
STSVD Root Directory
img_dir
with_smoke1.mp4
with_smoke2.mp4
... sequence ...
with_smoke120.mp4 120 files
clean_dir
without_smoke1.mp4
without_smoke120.mp4 120 files
ann_dir Annotations
diffusion_smoke
diff_smoke1.mp4 ... 80 files
ambient_smoke
amb_smoke41.mp4 ... 80 files
entangled_smoke
ent_smoke41.mp4 ... 40 files

Dataset Statistics

Dataset Statistics

Overview of the data generation pipeline and statistical distribution.

BibTeX

@misc{liang2025rethinkingsurgicalsmokesmoketypeaware,
      title={Rethinking Surgical Smoke: A Smoke-Type-Aware Laparoscopic Video Desmoking Method and Dataset}, 
      author={Qifan Liang and Junlin Li and Zhen Han and Xihao Wang and Zhongyuan Wang and Bin Mei},
      year={2025},
      eprint={2512.02780},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2512.02780}, 
}