SpecTf: Transformers Enable Data-Driven Imaging Spectroscopy Cloud Detection
Jake H. Lee, Michael Kiper, David R. Thompson, Philip G. Brodrick · 2025
Read SpecTf: Transformers Enable Data-Driven Imaging Spectroscopy Cloud Detection by Jake H. Lee, Michael Kiper, David R. Thompson, Philip G. Brodrick (2025) 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/arxiv-2501-04916.txt.
Description
Jake Lee, Michael Kiper, David Thompson, and Philip Brodrick built SpecTf in twenty twenty-five, a transformer that detects clouds in imaging spectroscopy data using spectra alone. Airborne and orbital spectrometers promise global measurement of Earth processes, but clouds corrupt their data, and classic screening leans on spatial patterns and repeat visits that variability and sparse coverage undermine. SpecTf treats each pixel's spectrum as a sequence to be read, not an image channel.
It significantly beat the operational baseline used for the EMIT instrument while matching heavier machine learning methods with far fewer parameters. Its attention maps expose physically meaningful spectral features it learned, making the model interpretable. Applied unchanged to a different instrument on a different platform, it still worked — evidence for instrument-agnostic algorithms in future spectroscopy missions.
Catalog record
| CREATOR | Jake H. Lee, Michael Kiper, David R. Thompson, Philip G. Brodrick |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 52 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-04916