An efficient reparameterized small object detection transformer for thermal infrared images
Canhao Guo, Peidong Luo, Zhixing Ma, Tianhong Zhao, Xiaole Shen · 2025
Read An efficient reparameterized small object detection transformer for thermal infrared images by Canhao Guo, Peidong Luo, Zhixing Ma, Tianhong Zhao, Xiaole Shen (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/pmc-12738694.txt.
Description
Canhao Guo and co-authors build a lean transformer that spots tiny objects in thermal infrared images from drones. Heat images have low contrast and little texture. Small targets vanish easily in such scenes.
Their model trains with many branches, then merges them for use. That trick cuts computing cost without losing quality. A new neck design mixes features across scales cheaply. Wavelet steps widen the view while keeping fine detail. Pruning trims the weakest weights by size afterward. The system targets edge devices on small aircraft. Tests show strong detection at modest compute budgets. The design suits search, patrol, and farming drones in the field.
Catalog record
| CREATOR | Canhao Guo, Peidong Luo, Zhixing Ma, Tianhong Zhao, Xiaole Shen |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 64 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |