CT-PestNet: A novel crop pest classification model based on cross-scale feature distillation enhancement and vision transformer with local token interaction
Youming Li, Weilin Cao, Liqiang Zhang · 2026
Read CT-PestNet: A novel crop pest classification model based on cross-scale feature distillation enhancement and vision transformer with local token interaction by Youming Li, Weilin Cao, Liqiang Zhang (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-0356061.txt.
Description
Youming Li, Weilin Cao, and Liqiang Zhang design a neural network that can tell crop pests apart in field photographs, in this 2026 paper. Pest losses eat a significant share of the world's food, and quick identification drives the right control response. Convolutional networks excel at fine local detail but miss long-range patterns, while vision transformers do the reverse. Their model, CT-PestNet, fuses both, adding a module that distills and boosts features across scales and another that lets neighboring image tokens interact more deeply.
On a large, difficult benchmark of more than one hundred pest classes, the model reached accuracy several points above the best prior convolutional and transformer models. It scored near-perfectly on an easier dataset. The authors credit the cross-scale treatment for handling pests of wildly different sizes against cluttered backgrounds, and present the system as practical support for field scouting.
Catalog record
| CREATOR | Youming Li, Weilin Cao, Liqiang Zhang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 45753 |
| LICENSE | CC_BY |
| RIGHTS | CC BY |
| SOURCE | PLOS |