YOLOv12-VSD: A Transfer-Learning-Assisted Real-Time Detection Algorithm for Vehicle Surface Defects
Haopu Liu, Dequn Zhao, Yu Li · 2026
Read YOLOv12-VSD: A Transfer-Learning-Assisted Real-Time Detection Algorithm for Vehicle Surface Defects by Haopu Liu, Dequn Zhao, Yu 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/pmc-13165786.txt.
Description
Haopu Liu, Dequn Zhao, and Yu Li propose a faster way to spot flaws on car surfaces in a 2026 paper. Finding dents and marks is hard because small flaws vary in size, scores can clash with exact spot data, and each factory line looks a bit different. Their method builds on a real time vision model called YOLOv12.
The new model, YOLOv12-VSD, adds a loss that ties scores to spot quality, a rebuilt block for richer mid level detail, and a wider pooling step for large flaws. A three stage transfer plan helps it adapt to new lines with few labels. In tests it beat six other models on the main score, while staying small and fast. Transfer raised its score and cut training time by 64 percent.
Catalog record
| CREATOR | Haopu Liu, Dequn Zhao, Yu Li |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 42 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |