Weighted multi-scale and wavelet-enhanced Segment Anything Model for salient object detection
Zhe Liu, Dan Tian · 2026
Read Weighted multi-scale and wavelet-enhanced Segment Anything Model for salient object detection by Zhe Liu, Dan Tian (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/plos-pone-0355742.txt.
Description
Zhe Liu and Dan Tian improved a well-known image segmentation model so it can pick out the most noticeable object in a scene. Adapting the general Segment Anything Model to saliency detection often leaves broken masks, fuzzy edges, and lost detail, especially when objects vary in size. Their 2026 design attacks three named weaknesses of earlier adapters.
A weighted multi-scale adapter learns how to calibrate features from different viewing scales instead of using fixed rules. A cross-fusion module ties levels together by meaning, not mere stacking. Wavelet processing rescues the fine, high-frequency detail that ordinary operations discard. Liu and Tian argue the three fixes together sharpen saliency maps for complex, textured scenes.
Catalog record
| CREATOR | Zhe Liu, Dan Tian |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 38182 |
| LICENSE | CC_BY |
| RIGHTS | CC BY |
| SOURCE | PLOS |