Object Counting on LOOKALIKES (test)
10MAED’Alessandro et al.
Evaluation results
| Method | Links | ||
|---|---|---|---|
| D’Alessandro et al.Protocol=per category synthetic data generation + adaptation2026.02 | 10 | 16.95 | |
| CountExProtocol=Trained on CoCount before zero-shot evaluation2026.02 | 18.53 | 30.46 | |
| CountGDProtocol=Zero-shot - no training/adaptation2026.02 | 22.34 | 33.9 | |
| LLMDetProtocol=Trained on CoCount before zero-shot evaluation2026.02 | 22.86 | 33.29 | |
| CountGDProtocol=Trained on CoCount before zero-shot evaluation2026.02 | 23.55 | 36.89 | |
| GroundingRECProtocol=Trained on CoCount before zero-shot evaluation2026.02 | 27.24 | 41.57 | |
| CAD-GDProtocol=Trained on CoCount before zero-shot evaluation2026.02 | 28.14 | 41.3 | |
| GroundingDINOProtocol=Zero-shot - no training/adaptation2026.02 | 33.89 | 59.51 | |
| OWLv2Protocol=Zero-shot - no training/adaptation2026.02 | 37.25 | 55.01 | |
| DAVEProtocol=Zero-shot - no training/adaptation2026.02 | 56.1 | 73.34 | |
| PseCoProtocol=Zero-shot - no training/adaptation2026.02 | 59.82 | 72.79 |