Depth-Sensitive Soft Suppression with RGB-D Inter-Modal Stylization Flow for Domain Generalization Semantic Segmentation

Binbin Wei, Yuhang Zhang, Shishun Tian, Muxin Liao, Wei Li, Wenbin Zou · 2025

Read Depth-Sensitive Soft Suppression with RGB-D Inter-Modal Stylization Flow for Domain Generalization Semantic Segmentation by Binbin Wei, Yuhang Zhang, Shishun Tian, Muxin Liao, Wei Li, Wenbin Zou (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-07050.txt.

Description

Binbin Wei and five co-authors present a study from twenty twenty-five on teaching vision models to label each part of an image. The hard case is a new scene the model has not seen in training. Depth maps, which show how far things are, can help a model cope with new scenes. But depth maps from cheap sensors have noise and holes.

Their method first restyles depth maps with help from the color image. It then mutes the parts of the depth map that react too strongly to style shifts. A soft match loss keeps the restyled depth close to the color data in part, while depth facts stay. Tests on many base networks show clear gains in scores for new scenes.

Catalog record

CREATORBinbin Wei, Yuhang Zhang, Shishun Tian, Muxin Liao, Wei Li, Wenbin Zou
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE47 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-07050

Browse the full library · About the Tower of Babel