Predictive Video Object Segmentation on FeVOS
39.5Jaccard Index (J)FeVOS-R1
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| FeVOS-R1Backbone=InternVL2.5-4B, Evaluation Protocol=Fine-tuned on FeVOS2026.06 | 39.5 | 45.1 | 42.3 | |
| Sa2VA*Backbone=InternVL2.5-4B, Evaluation Protocol=Fine-tuned on FeVOS2026.06 | 33.1 | 38.4 | 35.8 | |
| GLUS*Backbone=Chat-UniVi-7B, Evaluation Protocol=Fine-tuned on FeVOS2026.06 | 31 | 35.9 | 33.5 | |
| VRS-HQBackbone=Chat-UniVi-7B, Evaluation Protocol=Zero-shot2026.06 | 28.8 | 33.3 | 31 | |
| GLUSBackbone=Chat-UniVi-7B, Evaluation Protocol=Zero-shot2026.06 | 27.4 | 31.7 | 29.6 | |
| Sa2VABackbone=InternVL2.5-4B, Evaluation Protocol=Zero-shot2026.06 | 23.7 | 27.2 | 25.4 | |
| VISABackbone=Chat-UniVi-7B, Evaluation Protocol=Zero-shot2026.06 | 22.9 | 28.2 | 25.6 | |
| VideoLISABackbone=LLaVA-Phi-3-V-3.8B, Evaluation Protocol=Zero-shot2026.06 | 22.9 | 29.4 | 26.1 | |
| VideoGLaMMBackbone=Phi3-Mini-3.8B, Evaluation Protocol=Zero-shot2026.06 | 21.7 | 26.7 | 24.2 | |
| LMPMBackbone=Swin-T, Evaluation Protocol=Zero-shot2026.06 | 17.1 | 20.7 | 18.9 | |
| ReferFormerBackbone=ResNet-50, Evaluation Protocol=Zero-shot2026.06 | 16.4 | 20 | 18.2 |