HPA-UNet-LSNet: An LSNet-based U-Net with hybrid pooling attention for accurate segmentation of Haloxylon ammodendron crowns from UAV RGB imagery
Dongze Li, Xuefeng Yang, Yingnan Li, Ting Liang · 2026
Read HPA-UNet-LSNet: An LSNet-based U-Net with hybrid pooling attention for accurate segmentation of Haloxylon ammodendron crowns from UAV RGB imagery by Dongze Li, Xuefeng Yang, Yingnan Li, Ting Liang (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-0350455.txt.
Description
Counting saxaul shrubs across deserts helps track land restoration, and drones offer a bird's eye view, but the crowns blend into sand. Dongze Li and colleagues built a sharper image model for outlining these crowns in drone photos, in this 2026 PLOS One study.
Their design swaps the usual image backbone for a leaner one and adds an attention step that mixes pooled features. On held-out test images it marked crowns more precisely than the standard model, with far fewer false alarms and misses. Small crowns, the hardest cases, improved most. Heat maps showed the model focusing on crowns rather than background, a practical tool for desert monitoring.
Catalog record
| CREATOR | Dongze Li, Xuefeng Yang, Yingnan Li, Ting Liang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 57229 |
| LICENSE | CC_BY |
| RIGHTS | CC BY |
| SOURCE | PLOS |