LYEU-Net: A lightweight yet efficient U-Net with multibranch feature aggregation for pavement crack segmentation
Bing Yang, Jiaying He, Yuncheng Shen · 2026
Read LYEU-Net: A lightweight yet efficient U-Net with multibranch feature aggregation for pavement crack segmentation by Bing Yang, Jiaying He, Yuncheng Shen (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-0356870.txt.
Description
Bing Yang, Jiaying He, and Yuncheng Shen designed a pavement crack mapping network small enough to run on a phone. It still had to be accurate enough for road maintenance crews. Their 2026 paper rebuilds the classic U-Net design with three additions. A feature module exploits skip connections. A multibranch module reads cracks at several scales at once. An aggregation module weights features adaptively before the final outline is drawn.
Across four public crack datasets, the model posted the best overlap scores among rivals. It used a tiny fraction of their computing power. It traded a little recall for precision on some sets, missing faint cracks rather than inventing them. The authors released code and data, aiming at real-time inspection from ordinary mobile devices.
Catalog record
| CREATOR | Bing Yang, Jiaying He, Yuncheng Shen |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 48645 |
| LICENSE | CC_BY |
| RIGHTS | CC BY |
| SOURCE | PLOS |