A fine-grained defect-sensitive local–global multiscale vision transformer framework for urban visual pollution prediction
Peng Xu, Jingyi Shen, Taotao Zou, Fa Shen, Yalun Lei, Digong Liu · 2026
Read A fine-grained defect-sensitive local–global multiscale vision transformer framework for urban visual pollution prediction by Peng Xu, Jingyi Shen, Taotao Zou, Fa Shen, Yalun Lei, Digong Liu (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/PMC13493999.txt.
Description
Peng Xu and five colleagues present in 2026 a vision system that grades how visually polluted a city street looks, from garbage piles to graffiti to potholes. Manual inspections are slow and subjective, so their transformer model reads street level photographs through both fine grained local patches and whole scene context, hunting small defects generic models miss. A built-in index converts detections into a single severity number.
Across seven pollution categories the system scored about ninety-five percent accuracy, edging out baseline transformers on precision and recall alike. Its severity index agreed closely with its own classifications. Four different explanation techniques confirmed the model stares at the actual mess rather than background scenery. City agencies could survey whole districts from camera cars instead of clipboards.
Catalog record
| CREATOR | Peng Xu, Jingyi Shen, Taotao Zou, Fa Shen, Yalun Lei, Digong Liu |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 133326 |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |