Semi-supervised Video Object Segmentation on SA-V (val)
78.6J&F ScoreSAM 2
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SAM 2Backbone=Hiera-L, Training Data=SA-1B, SA-V, and internal2024.08 | 78.6 | 75.3 | 82 | |
| SAM 2Backbone=Hiera-L2024.08 | 77.9 | 74.5 | 81.3 | |
| SAM 2Backbone=Hiera-B+, Training Data=SA-1B, SA-V, and internal2024.08 | 77.5 | 74.1 | 80.9 | |
| SAM 2Backbone=Hiera-S, Training Data=SA-1B, SA-V, and internal2024.08 | 77 | 73.6 | 80.5 | |
| SAM 2Backbone=Hiera-B+2024.08 | 76.8 | 73.4 | 80.1 | |
| SAM 2Backbone=Hiera-T, Training Data=SA-1B, SA-V, and internal2024.08 | 75.2 | 71.7 | 78.7 | |
| SAM 2.1 (T)Parameters (M)=38.9, Training data ratio=1, Tokens ratio=1, FPS (V100)=8.82026.05 | 75.2 | — | — | |
| EdgeTAMParameters (M)=13.9, Training data ratio=0.92, Tokens ratio=0.14, FPS (V100)=19.22026.05 | 72.3 | — | — | |
| EfficientTAMParameters (M)=18.0, Training data ratio=0.92, Tokens ratio=0.25, FPS (V100)=20.12026.05 | 71.3 | — | — | |
| TinySAM 2Parameters (M)=17.5, Training data ratio=0.03, Tokens ratio=0.07, FPS (V100)=25.62026.05 | 65.1 | — | — | |
| SwinB-DeAOT-L2024.08 | 61.4 | 56.6 | 66.2 | |
| Cutie-base+2024.08 | 61.3 | 58.3 | 64.4 | |
| Cutie-base+Parameters (M)=35.0, Training data ratio=0.02, Tokens ratio=1, FPS (V100)=17.92026.05 | 61.3 | — | — | |
| STCN2024.08 | 61 | 57.4 | 64.5 | |
| STCNParameters (M)=54.0, Training data ratio=0.02, Tokens ratio=1, FPS (V100)=13.22026.05 | 61 | — | — | |
| Cutie-base2024.08 | 60.7 | 57.7 | 63.7 | |
| Cutie-baseParameters (M)=35.0, Training data ratio=0.02, Tokens ratio=12026.05 | 60.7 | — | — | |
| XMem2024.08 | 60.1 | 56.3 | 63.9 | |
| XMemParameters (M)=62.0, Training data ratio=0.02, Tokens ratio=12026.05 | 60.1 | — | — | |
| DEVA2024.08 | 55.4 | 51.5 | 59.2 | |
| DEVAParameters (M)=69.0, Training data ratio=0.02, Tokens ratio=12026.05 | 55.4 | — | — | |
| SAM-I2VParameters (M)=18.9, Training data ratio=0.02, Tokens ratio=1, FPS (V100)=6.62026.05 | 55.4 | — | — | |
| RDE2024.08 | 51.8 | 48.4 | 55.2 | |
| RDEParameters (M)=64.0, Training data ratio=0.02, Tokens ratio=12026.05 | 51.8 | — | — | |
| SwinB-AOT-L2024.08 | 51.1 | 46.4 | 55.7 | |
| SimVOS-B2024.08 | 44.2 | 40 | 48.3 |