Abstract
Smoke in real-world scenes can severely degrade image quality and hamper visibility. Recent image restoration methods either rely on data-driven priors that are susceptible to hallucinations, or are limited to static low-density smoke. We introduce SmokeSeer, a method for simultaneous 3D scene reconstruction and smoke removal from multi-view video sequences. Our method uses thermal and RGB images, leveraging the reduced scattering in thermal images to see through smoke. We build upon 3D Gaussian splatting to fuse information from the two image modalities, and decompose the scene into smoke and non-smoke components. Unlike prior work, SmokeSeer handles a broad range of smoke densities and adapts to temporally varying smoke. We validate our method on synthetic data and a new real-world smoke dataset with RGB and thermal images.
Method
Results
Resources
Paper: Our paper is available on arXiv and locally.
Code: Our code is available on Github.
Data: Our data is available on Google Drive.
Citation
@inproceedings{jain2025smokeseer,
title={{SmokeSeer: 3D Gaussian Splatting for Smoke Removal and Scene Reconstruction}},
author={Neham Jain and Andrew Jong and Sebastian Scherer and Ioannis Gkioulekas},
booktitle={International Conference on 3D Vision (3DV)},
year={2026},
publisher={IEEE},
organization={IEEE},
}
Acknowledgments
We thank Ian Higgins and John Keller for help with flying the drone during data acquisition; Sreekar Ranganathan for help with the camera hardware and data collection setup; and Jeff Tan and Nikhil Keetha for helpful discussions. This work was supported by National Institute of Food and Agriculture award 2023-67021-39073; Defense Science and Technology Agency contract #DST000EC124000205; and Alfred P. Sloan Research Fellowship FG202013153 for Ioannis Gkioulekas. This work used Bridges-2 at the Pittsburgh Supercomputing Center, through allocation cis220039p from the Advanced Cyberinfrastructure Coordination Ecosystem: Services & Support (ACCESS) program, which is supported by National Science Foundation awards 2138259, 2138286, 2138307, 2137603, and 2138296; and National Artificial Intelligence Research Resource Pilot-2211.