Object Discovery on MOVi-D
45.3ARISlotContrast
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| SlotContrast#slot=21, conditional=true, Input resolution=256×256 (224×224), Encoder=DINO2 ViT-S/142025.08 | 45.3 | 63.9 | 26.7 | 25.1 | |
| RandSF.Q#slot=21, conditional=true, loss=ssc (slot-slot contrast), Input resolution=256×256 (224×224), Encoder=DINO2 ViT-S/142025.08 | 41.6 | 77.5 | 27.4 | 25.6 | |
| RandSF.Q#slot=21, conditional=true, loss=tsim (time similarity), Input resolution=256×256 (224×224), Encoder=DINO2 ViT-S/142025.08 | 41.2 | 72 | 27.1 | 25.4 | |
| STEVE#slot=21, conditional=true, Input resolution=256×256 (224×224), Encoder=DINO2 ViT-S/142025.08 | 32.7 | 66.5 | 23 | 21.2 | |
| VideoSAUR#slot=21, conditional=true, Input resolution=256×256 (224×224), Encoder=DINO2 ViT-S/142025.08 | 22.5 | 40 | 11.6 | 10.8 |