MSPFormer: An enhanced multi-scale and semantic-preserving transformer for sheep ownership identification in precision livestock farming
Yaosheng Han, Chunmei Li, Xiangjie Huang, Qing Dong, Hao Wang · 2026
Read MSPFormer: An enhanced multi-scale and semantic-preserving transformer for sheep ownership identification in precision livestock farming by Yaosheng Han, Chunmei Li, Xiangjie Huang, Qing Dong, Hao Wang (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-0349654.txt.
Description
Han and colleagues, in a 2026 paper, build a vision model that can tell whose sheep is whose on the Qinghai-Tibet Plateau. Overgrazing is wearing down the grassland, and knowing which flock grazes where would help balance herds with grass and settle disputes between herders. Sheep carry natural color patterns on their backs, and the team trained a segmentation model to read them.
Their model upgrades a strong existing design with three additions. A module reads the scene at several scales. Another keeps meaning consistent. A third rebuilds fine detail when enlarging the output. On a purpose-built data set of sheep in varied light and backgrounds, it beat its base model on every score. It also held up on a public benchmark and small test sets. The authors offer the tool for precision grazing management on fragile high grassland.
Catalog record
| CREATOR | Yaosheng Han, Chunmei Li, Xiangjie Huang, Qing Dong, Hao Wang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 102058 |
| LICENSE | CC_BY |
| RIGHTS | CC BY |
| SOURCE | PLOS |