Referring Video Object Segmentation on Ref-Youtube-VOS 2019 (test)
64.9J&F ScoreReferFormer
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ReferFormerBackbone=Video-Swin-B, joint training with Ref-COCO=true2022.01 | 64.9 | 62.8 | 67 | |
| ReferFormerBackbone=Swin-L, joint training with Ref-COCO=true2022.01 | 64.2 | 62.3 | 66.2 | |
| ReferFormerBackbone=Video-Swin-S, joint training with Ref-COCO=true2022.01 | 63.3 | 61.4 | 65.2 | |
| ReferFormerBackbone=Video-Swin-B2022.01 | 62.9 | 61.3 | 64.6 | |
| ReferFormerBackbone=Video-Swin-T, joint training with Ref-COCO=true2022.01 | 62.6 | 59.9 | 63.3 | |
| ReferFormerBackbone=Swin-L2022.01 | 62.4 | 60.8 | 64 | |
| CITDBackbone=Ensemble2022.01 | 61.4 | 60 | 62.7 | |
| ReferFormerBackbone=Video-Swin-S2022.01 | 60.1 | 58.6 | 61.6 | |
| ReferFormerBackbone=Video-Swin-T2022.01 | 59.4 | 58 | 60.9 | |
| ReferFormerBackbone=ResNet-101, joint training with Ref-COCO=true2022.01 | 59.3 | 58.1 | 60.4 | |
| ReferFormerBackbone=ResNet-50, joint training with Ref-COCO=true2022.01 | 58.7 | 57.4 | 60.1 | |
| ReferFormerBackbone=ResNet-1012022.01 | 57.3 | 56.1 | 58.4 | |
| CITDBackbone=ResNet-101, joint training with Ref-COCO=true2022.01 | 56.4 | 54.8 | 58.1 | |
| ReferFormerBackbone=Video-Swin-T, window size=w = 5, trained from scratch=true2022.01 | 56 | 54.8 | 57.3 | |
| ReferFormerBackbone=ResNet-502022.01 | 55.6 | 54.8 | 56.5 | |
| MTTRwindow size=w = 12, Visual Backbone Type=Spatio-temporal2022.01 | 55.3 | 54 | 56.6 | |
| PMINet + CFBIBackbone=Ensemble2022.01 | 54.2 | 53 | 55.5 | |
| PMINet + CFBIBackbone=ResNeSt-1012022.01 | 53 | 51.5 | 54.5 | |
| PMINetBackbone=ResNeSt-1012022.01 | 48.2 | 46.7 | 49.6 | |
| URVOSBackbone=ResNet-502022.01 | 47.2 | 45.3 | 49.2 | |
| CMSA + RNNBackbone=ResNet-502022.01 | 36.4 | 34.8 | 38.1 | |
| CMSABackbone=ResNet-502022.01 | 34.9 | 33.3 | 36.5 |