Dataset Grouping on COCO 4c
62.6Activity CAccCLIP
Evaluation Results
| Method | Links | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| CLIPmode=Zero-shot2024.10 | 62.6 | — | — | 73.5 | 67.6 | 34.3 | 51.5 | 41.1 | 22.4 | 43.3 | 29.5 | 40.6 | 74.1 | 52.4 | |
| BLIP-2Type=Img-based2024.10 | 48.7 | — | — | 64.1 | 55.3 | 39.6 | 48 | 43.4 | 30.2 | 37.5 | 33.4 | 40.7 | 60.3 | 48.6 | |
| LLaVAType=Img-based2024.10 | 46.5 | — | — | 61.8 | 53.1 | 34 | 46.3 | 39.2 | 28 | 24.7 | 26.3 | 39.4 | 51.7 | 44.7 | |
| X-ClusterInput=Caption-based2024.10 | 44.1 | — | — | 48.9 | 46.4 | 55.2 | 55.6 | 55.4 | 38.1 | 32.6 | 35.2 | 67.6 | 56.7 | 61.7 | |
| Tag-basedInput=Tags2024.10 | 43.2 | — | — | 51.5 | 47 | 28.6 | 46.6 | 35.5 | 13 | 25.6 | 17.2 | 19.3 | 48.8 | 27.7 | |
| KMeans DINOv2Backbone=DINOv22024.10 | 38.2 | — | — | — | — | 37.9 | — | — | 22.5 | — | — | 43.8 | — | — | |
| KMeans DINOv1Backbone=DINOv12024.10 | 34.8 | — | — | — | — | 37.5 | — | — | 17.9 | — | — | 36.5 | — | — | |
| KMeans CLIPBackbone=CLIP2024.10 | 34.4 | — | — | — | — | 32.7 | — | — | 18.9 | — | — | 38.6 | — | — | |
| IC|TCprior_input=Ground-truth criteria and cluster count (Kl)2024.10 | — | 48.9 | 53.2 | — | — | — | — | — | — | — | — | — | — | — | |
| MMaPprior_input=Ground-truth criteria and cluster count (Kl)2024.10 | — | 33.9 | — | — | — | — | — | — | — | — | — | — | — | — | |
| MSubprior_input=Ground-truth criteria and cluster count (Kl)2024.10 | — | 36 | — | — | — | — | — | — | — | — | — | — | — | — | |
| SSD-LLMprior_input=Ground-truth criteria and cluster count (Kl)2024.10 | — | 41.6 | 52.1 | — | — | — | — | — | — | — | — | — | — | — | |
| UBmode=Zero-shot, backbone=CLIP L/142024.10 | — | 40.1 | 60.6 | — | — | — | — | — | — | — | — | — | — | — | |
| X-Clusterdiscovery=Fully automated2024.10 | — | 51.2 | 48.4 | — | — | — | — | — | — | — | — | — | — | — |