Few-Shot Video Instance Segmentation on YouTube-VIS
48.4mIoU (Fold 1)TTI
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| TTIBackbone=ResNet-50, Support set=five-shot one-way, Episodic training=false, Whole video treatment=true2022.03 | 48.4 | 68.5 | 62.6 | 62.4 | 60.5 | 57.7 | 81.6 | 73.6 | 75.6 | 72.1 | |
| RePRIBackbone=ResNet-50, Support set=five-shot one-way, Episodic training=false, Whole video treatment=false2022.03 | 45.8 | 68.6 | 59.3 | 64.2 | 59.5 | 54.1 | 75.6 | 63.9 | 71.6 | 66.3 | |
| PPNetBackbone=ResNet-50, Support set=five-shot one-way, Episodic training=true, Whole video treatment=false2022.03 | 45.5 | 63.8 | 60.4 | 58.9 | 57.1 | — | — | — | — | — | |
| DANetBackbone=ResNet-50, Support set=five-shot one-way, Episodic training=true, Whole video treatment=true2022.03 | 43.2 | 65 | 62 | 61.8 | 58 | 32.3 | 63.7 | 57.2 | 58 | 52.8 | |
| PFENetBackbone=ResNet-50, Support set=five-shot one-way, Episodic training=true, Whole video treatment=false2022.03 | 37.8 | 64.4 | 56.3 | 56.4 | 53.7 | — | — | — | — | — | |
| Naive Temporal RePRIBackbone=ResNet-50, Support set=five-shot one-way, Episodic training=false, Whole video treatment=true2022.03 | 36.6 | 62 | 50.2 | 55.2 | 51 | 36 | 57.3 | 49.1 | 53.2 | 48.9 | |
| PMMsBackbone=ResNet-50, Support set=five-shot one-way, Episodic training=true, Whole video treatment=false2022.03 | 32.9 | 61.1 | 56.8 | 55.9 | 51.7 | — | — | — | — | — |