Text-to-Thermal Retrieval (via Vision pivot) on LLVIP to FLIR (test)
37.6mAPIB+BB
Evaluation Results
| Method | Links | |
|---|---|---|
| IB+BBBackbone=ImageBind, Variant=BrokenBind2026.02 | 37.6 | |
| LB+BBBackbone=LanguageBind, Variant=BrokenBind2026.02 | 36.8 | |
| LB+FTBackbone=LanguageBind, Variant=Fine-tuned2026.02 | 19.9 | |
| IB+FTBackbone=ImageBind, Variant=Fine-tuned2026.02 | 18.9 | |
| LBBackbone=LanguageBind, Variant=Zero-shot2026.02 | 12.5 | |
| IBBackbone=ImageBind, Variant=Zero-shot2026.02 | 11.6 |