Explainable Superpixel-Guided Graph Vision Transformer for Hyperspectral Image Analysis
Jieli Chen, Kah Phooi Seng, Chee Shen Lim, Li-Minn Ang, Jeremy Smith · 2026
Read Explainable Superpixel-Guided Graph Vision Transformer for Hyperspectral Image Analysis by Jieli Chen, Kah Phooi Seng, Chee Shen Lim, Li-Minn Ang, Jeremy Smith (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-13568120.txt.
Description
Jieli Chen, Kah Phooi Seng, Chee Shen Lim, and co-authors present a vision model that both classifies satellite images and explains itself in this 2026 paper. Hyperspectral images record many wavelengths per pixel, which separates materials finely. Land regions have ragged shapes, though, and fixed square patches fit them badly.
Their model runs two branches. One groups pixels into adaptive superpixels and treats them as a graph of tokens for a transformer. The other keeps fine local detail with windowed attention. The two views are fused for the final call on each pixel. From its own gradients the model draws maps of which regions and which wavelengths drove each class. Benchmark tests showed better accuracy than patch-based rivals, with clearer explanations. The authors aim the tool at land mapping where users need reasons, not just labels.
Catalog record
| CREATOR | Jieli Chen, Kah Phooi Seng, Chee Shen Lim, Li-Minn Ang, Jeremy Smith |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 59 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |