Spectral Rectification for Parameter-Efficient Adaptation of Foundation Models in Colonoscopy Depth Estimation Journal: Medical Image Analysis
Xiaoxian Zhang, Minghai Shi, Lei Li · 2026
Read Spectral Rectification for Parameter-Efficient Adaptation of Foundation Models in Colonoscopy Depth Estimation Journal: Medical Image Analysis by Xiaoxian Zhang, Minghai Shi, Lei Li (2026) 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-2603-15374.txt.
Description
Zhang and co-authors present SpecDepth, a light adaptation of vision foundation models for depth estimation in colonoscopy. Models trained on natural images fail here, not from a meaning gap but from a frequency shift. Colonoscopy images lack the sharp edges and textures the models lean on. A learnable wavelet module boosts the faded high-frequency parts of feature maps. This low-level fix keeps the pretrained geometry skills intact. On C3VD and SimCol3D, the method set state-of-the-art scores with very few added parameters.
Catalog record
| CREATOR | Xiaoxian Zhang, Minghai Shi, Lei Li |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 59832 |
| LICENSE | http://creativecommons.org/licenses/by-sa/4.0/ |
| RIGHTS | CC BY SA |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.15374