Thermal-Det: Language-Guided Cross-Modal Distillation for Open-Vocabulary Thermal Object Detection
Yasiru Ranasinghe, Elim Schenck, Florence Yellin, Shuowen Hu, Christopher Funk, Vishal M. Patel · 2026
Read Thermal-Det: Language-Guided Cross-Modal Distillation for Open-Vocabulary Thermal Object Detection by Yasiru Ranasinghe, Elim Schenck, Florence Yellin, Shuowen Hu, Christopher Funk, Vishal M. Patel (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/arxiv-2605-10130.txt.
Description
Ranasinghe and five co-authors build the first large-language-model-supervised open-vocabulary tool designed for thermal images. Existing open-vocabulary detectors target RGB photos and fail on thermal imagery, where low texture and emissivity variation defeat RGB-based meanings. Thermal-Det is trained on a made-up set made by turning GroundingCap-1M into the thermal domain and sifting out RGB-specific caption terms, yielding over one million aligned samples.
The model jointly tunes spotting, caption work and cross-modal distillation, with a frozen RGB teacher supplying shape and meaning pseudo-guidance for paired but unlabelled RGB-thermal data. A Thermal-Text Alignment Head calibrates text, and a Modality-Fused Cross-Attention module handles dual-modality reasoning. Fully fine-tuned to internalise thermal contrast patterns while preserving language alignment, it gains a consistent two to four percent average precision over existing open-vocabulary tools on public benchmarks.
Catalog record
| CREATOR | Yasiru Ranasinghe, Elim Schenck, Florence Yellin, Shuowen Hu, Christopher Funk, Vishal M. Patel |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 52 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.10130