Leveraging generative adversarial networks and SE-UNet for high-precision tobacco leaf image segmentation and blend uniformity detection
Meizhou Ding, Xianhui Peng, Xiaonan Ji, Song Chen, Quansheng Li, Helan Zou, Zhiwei Fan, Shucai Wang · 2026
Read Leveraging generative adversarial networks and SE-UNet for high-precision tobacco leaf image segmentation and blend uniformity detection by Meizhou Ding, Xianhui Peng, Xiaonan Ji, Song Chen, Quansheng Li, Helan Zou, Zhiwei Fan, Shucai 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/PMC13572454.txt.
Description
Meizhou Ding, with seven colleagues, presents a 2026 method for checking how evenly tobacco leaf types are mixed, a factor that shapes cigarette quality. Hand sampling is slow and may miss how the blend looks across a full batch. The team built a real-time image system that first creates more training images, then outlines each leaf type in live pictures.
Three image-making networks were compared, and CycleGAN gave the best leaf images. An attention step was added to a U-Net model so it could stress key features and mute background noise. The tuned model beat four rival models on overlap scores, with gains from 2.77 to 17.46 points. Pixel counts for each leaf type then gave the blend share in each image, a route the authors say could serve other farm goods too.
Catalog record
| CREATOR | Meizhou Ding, Xianhui Peng, Xiaonan Ji, Song Chen, Quansheng Li, Helan Zou, Zhiwei Fan, Shucai Wang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49780 |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |