Action Retrieval on SCLARO
57.63Top-1 AccuracyScenarioCLIP
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ScenarioCLIPmode=Zero-shot, pretraining=In-domain (SCLARO), retrieval_type=image→text2025.11 | 57.63 | 75.02 | 79.99 | |
| ScenarioCLIP w/o KDmode=Zero-shot, pretraining=In-domain (SCLARO), retrieval_type=image→text2025.11 | 57.52 | 75.24 | 80.34 | |
| PyramidCLIPmode=Zero-shot, pretraining=In-domain (SCLARO), retrieval_type=image→text2025.11 | 54.95 | 73.36 | 78.84 | |
| CLIPmode=Zero-shot, pretraining=released weights (prompt-only), retrieval_type=image→text2025.11 | 14.87 | 33.43 | 42.45 | |
| FineCLIPmode=Zero-shot, pretraining=released weights (prompt-only), retrieval_type=image→text2025.11 | 14.76 | 32.88 | 42 | |
| SLIPmode=Zero-shot, pretraining=released weights (prompt-only), retrieval_type=image→text2025.11 | 8.31 | 21.83 | 30.26 | |
| TripletCLIPmode=Zero-shot, pretraining=released weights (prompt-only), retrieval_type=image→text2025.11 | 5.48 | 14.91 | 21.67 |