Image-to-Text Retrieval on FLIR (test)
12.3Recall@1T-CLIP
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| T-CLIPSetting=Dual2026.05 | 12.3 | 35.7 | 51.7 | 72.7 | 85.1 | 93.1 | |
| T-CLIPSetting=Global2026.05 | 11.9 | 35.3 | 50.4 | 70.7 | 83.3 | 92.3 | |
| Global LoRA2026.05 | 10.5 | 32.9 | 48.7 | 67.8 | 81.6 | 91.4 | |
| DeCLIPSetting=Global2026.05 | 4.9 | 15.4 | 22.8 | 37.6 | 50.7 | 64.2 | |
| CLIP-AdapterSetting=Global2026.05 | 4.5 | 15.3 | 24.7 | 41.1 | 54.5 | 70 | |
| CLIP-AdapterSetting=Fine-grained2026.05 | 2.7 | 8.1 | 13.4 | 25 | 40.2 | 58.2 | |
| F-G LoRASetting=Fine-grained2026.05 | 2.7 | 8.9 | 15 | 29 | 45.3 | 64.2 | |
| T-CLIPSetting=Fine-grained2026.05 | 2.5 | 10.7 | 18.2 | 31.6 | 47.3 | 64.9 | |
| CLIPSetting=Global2026.05 | 1.4 | 7 | 11.5 | 24.2 | 36 | 51.3 | |
| DeCLIPSetting=Fine-grained2026.05 | 1.2 | 5.6 | 10.4 | 19.7 | 31.2 | 47.4 | |
| CLIPSetting=Fine-grained2026.05 | 0.6 | 2.2 | 3.9 | 11 | 22 | 33.2 |