This file documents a collection of models reported in our paper: Learning Long-term Visual Dynamics with Region Proposal Interaction Networks. Links to the pretrained models are provided as well.
We use T=5 for PHYRE, T=15 for ShapeStacks, and T=20 for training SimB/RealB:
Dataset | Error @ [0, T] | Error @ [T, 2 * T] | download |
---|---|---|---|
PHYRE (Within) | 1.308 | 11.060 | model |
PHYRE (Cross) | 3.262 | 15.185 | model |
ShapeStacks (3) | 1.027 | 4.728 | model |
ShapeStacks (4) | 5.753 | 23.459 | model |
RealB | 0.296 | 2.336 | model |
SimB | 2.55 | 25.77 | model |
Fold ID | AUCCESS | download |
---|---|---|
0 | 85.49 | model |
1 | 86.57 | model |
2 | 85.58 | model |
3 | 84.11 | model |
4 | 85.30 | model |
5 | 85.18 | model |
6 | 84.78 | model |
7 | 84.32 | model |
8 | 85.71 | model |
9 | 85.17 | model |
Fold ID | AUCCESS | download |
---|---|---|
0 | 50.86 | model |
1 | 36.58 | model |
2 | 55.44 | model |
3 | 38.34 | model |
4 | 37.11 | model |
5 | 47.23 | model |
6 | 38.23 | model |
7 | 47.19 | model |
8 | 32.23 | model |
9 | 38.76 | model |