Image Retrieval on OpenEvents v1 (public test)
0.559mAPLightweight Entity-Guided Event-Based Image Retrieval (Ours)
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Lightweight Entity-Guided Event-Based Image Retrieval (Ours)Backbone=BEiT-3, Filtering=Entity-driven (BM25)2025.12 | 0.559 | — | — | — | |
| SBERT + Bart + CLIPRetrieval=SBERT, Language Model=Bart, Vision Model=CLIP2025.12 | 0.3232 | 0.3978 | 0.2226 | 0.0436 | |
| SBERT + Pegasus + CLIPRetrieval=SBERT, Language Model=Pegasus, Vision Model=CLIP2025.12 | 0.3216 | 0.3986 | 0.2173 | 0.045 | |
| SBERT + PegasusRetrieval=SBERT, Language Model=Pegasus2025.12 | 0.2868 | 0.3665 | 0.193 | 0.0362 | |
| SBERT + BartRetrieval=SBERT, Language Model=Bart2025.12 | 0.284 | 0.3628 | 0.1863 | 0.0372 | |
| SBERT + FlanT5 + CLIPRetrieval=SBERT, Language Model=FlanT5, Vision Model=CLIP2025.12 | 0.2795 | 0.3408 | 0.1986 | 0.0303 | |
| CLIPModel Type=Vision-Language Model2025.12 | 0.2467 | 0.3407 | 0.1586 | 0.0302 | |
| SBERT + FlanT5Retrieval=SBERT, Language Model=FlanT52025.12 | 0.2134 | 0.2837 | 0.1376 | 0.022 | |
| OpenCLIPModel Type=Vision-Language Model2025.12 | 0.1845 | 0.2703 | 0.1845 | 0.0185 |