TUH-NAS: A Triple-Unit NAS Network for Hyperspectral Image Classification
Feng Chen, Baishun Su, Zongpu Jia · 2024
Read TUH-NAS: A Triple-Unit NAS Network for Hyperspectral Image Classification by Feng Chen, Baishun Su, Zongpu Jia (2024) 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-11644869.txt.
Description
Feng Chen, Baishun Su, and Zongpu Jia let a computer design the neural network that reads satellite images by their spectra. Hyperspectral images record hundreds of wavelengths per pixel, marrying rich spectral detail to spatial pattern, and hand-built networks rarely fuse the two well. Architecture search automates the design, but earlier searches ignored the spectral-spatial bond. Their 2024 paper builds a triple-unit search network around that bond.
One unit exists purely to fuse the two kinds of information. A new attention module aims the network at the image's critical regions, and a composite loss makes it study the samples it finds hardest. On three public datasets, with few training samples, the searched network beat earlier architecture-search classifiers, and it drew object boundaries more faithfully. The authors present the fusion-first design as the point: in hyperspectral work, the marriage of spectrum and space is the classification.
Catalog record
| CREATOR | Feng Chen, Baishun Su, Zongpu Jia |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 62 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |