Summary of Weathergs: 3d Scene Reconstruction in Adverse Weather Conditions Via Gaussian Splatting, by Chenghao Qian et al.
WeatherGS: 3D Scene Reconstruction in Adverse Weather Conditions via Gaussian Splatting
by Chenghao Qian, Yuhu Guo, Wenjing Li, Gustav Markkula
First submitted to arxiv on: 25 Dec 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The proposed WeatherGS framework addresses the challenge of reconstructing clear 3D scenes from multi-view images under different weather conditions. The framework builds upon 3D Gaussian Splatting (3DGS) and explicitly categorizes artifacts into dense particles caused by air-borne precipitation, and lens occlusions raised by precipitation on the camera lens. A dense-to-sparse preprocess strategy is employed to remove dense particles using an Atmospheric Effect Filter (AEF) and extract sparse occlusion masks with a Lens Effect Detector (LED). The framework trains 3D Gaussians based on processed images and generated masks for excluding occluded areas, and accurately recovers the underlying clear scene through Gaussian splatting. The WeatherGS is evaluated on a diverse benchmark demonstrating high-quality, clean scenes across various weather scenarios, outperforming existing state-of-the-art methods. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary WeatherGS is a new way to reconstruct 3D scenes from images taken in different weather conditions like snow or rain. Right now, 3D Gaussian Splatting (3DGS) has trouble with these types of scenes because it treats the artifacts caused by weather as part of the scene. This makes the reconstructed scene unclear. To fix this, WeatherGS categorizes the artifacts into two types: dense particles from air-borne precipitation and lens occlusions from precipitation on the camera lens. The framework then removes the dense particles and extracts sparse masks to exclude occluded areas and recover the underlying clear scene. |
