Image Retrieval on FSIR-BD-Compositional-VG 1.0
45.9mAPBLIP-L PL
Evaluation Results
| Method | Links | |
|---|---|---|
| BLIP-L PLImage Encoder=BLIP-L, Optimization Strategy=FSIR-PL2026.03 | 45.9 | |
| BLIP-LImage Encoder=BLIP-L, Optimization Strategy=Baseline2026.03 | 41 | |
| FSIR-CTR-Phi-3.5-BLIP-LImage Encoder=BLIP-L, Optimization Strategy=FSIR-CTR, Language Model=Phi-3.52026.03 | 40.8 | |
| FSIR-CTR-Phi-3.5-CLIP-LImage Encoder=CLIP-L, Optimization Strategy=FSIR-CTR, Language Model=Phi-3.52026.03 | 29.7 | |
| CLIP-L PLImage Encoder=CLIP-L, Optimization Strategy=FSIR-PL2026.03 | 28.3 | |
| CLIP-LImage Encoder=CLIP-L, Optimization Strategy=Baseline2026.03 | 25.9 |