Semantic Forecasting on Frozen Forecasting Unified Evaluation
0.28Mean Jaccard DistanceDINOv2
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| DINOv2Diffusion Model Samples=10, Context Frames=42025.07 | 0.28 | 0.066 | 0.39 | 0.49 | 37 | 0.057 | 46 | 52 | |
| SigLIPDiffusion Model Samples=10, Context Frames=42025.07 | 0.37 | 0.037 | 0.42 | 0.31 | 14 | 0.07 | 24 | 20 | |
| 4DS-hDiffusion Model Samples=10, Context Frames=42025.07 | 0.4 | 0.11 | 0.58 | 0.82 | 43 | 0.13 | 63 | 74 | |
| VideoPrismDiffusion Model Samples=10, Context Frames=42025.07 | 0.54 | 0.049 | 0.61 | 0.74 | 44 | 0.045 | 51 | 63 | |
| VideoMAEv2Diffusion Model Samples=10, Context Frames=42025.07 | 0.56 | 0.04 | 0.61 | 0.73 | 56 | 0.053 | 64 | 72 | |
| N-WALT 500MDiffusion Model Samples=10, Context Frames=42025.07 | 0.57 | 0.012 | 0.59 | 0.76 | 46 | 0.027 | 51 | 68 | |
| 4DS-eDiffusion Model Samples=10, Context Frames=42025.07 | 0.58 | 0.029 | 0.61 | 0.83 | 56 | 0.071 | 66 | 77 | |
| VJEPADiffusion Model Samples=10, Context Frames=42025.07 | 0.59 | 0.023 | 0.63 | 0.79 | 49 | 0.054 | 58 | 74 | |
| VideoMAEDiffusion Model Samples=10, Context Frames=42025.07 | 0.6 | 0.026 | 0.63 | 0.79 | 56 | 0.038 | 62 | 74 | |
| WALT 500MDiffusion Model Samples=10, Context Frames=42025.07 | 0.64 | 0.03 | 0.68 | 0.76 | 50 | 0.05 | 58 | 68 |