Unsupervised Video Object Segmentation on SegTrack v2
81.2Jaccard ScoreLRTL
Evaluation results
| Method | Links | |
|---|---|---|
| LRTLInf. Input RGB=true, Inf. Input Motion=false, Input Resolution=192 × 352, Motion Est. Method=RAFT [60] / CoTracker [28]2025.01 | 81.2 | |
| LOCATEInf. Input RGB=true, Inf. Input Motion=false, Input Resolution=480 × 848, Motion Est. Method=ARFlow [39]2025.01 | 79.9 | |
| RCFPost-process=CRF only2023.04 | 79.6 | |
| Ours (w/ CRF)extra post-processing=true2022.05 | 78.9 | |
| GWMInf. Input RGB=true, Inf. Input Motion=false, Input Resolution=128 × 224, Motion Est. Method=RAFT [60]2025.01 | 78.9 | |
| RCF (w/o SC)Post-process=CRF only2023.04 | 78.7 | |
| Ours (MaskFormer)Input Flow=false2022.05 | 78.3 | |
| Ours (UNet)Input Flow=false2022.05 | 76.8 | |
| RCFPost-process=None2023.04 | 76.7 | |
| RCFInf. Input RGB=true, Inf. Input Motion=false, Input Resolution=480 × 848, Motion Est. Method=RAFT [60]2025.01 | 76.7 | |
| Ponimatkin et al.Inf. Input RGB=true, Inf. Input Motion=false, Input Resolution=480 × 848, Motion Est. Method=ARFlow [39]2025.01 | 74.9 | |
| DyStaBextra post-processing=true, supervised pre-training=true2022.05 | 74.2 | |
| OCLR (test ft.)Inf. Input RGB=true, Inf. Input Motion=true, Input Resolution=480 × 848, Motion Est. Method=RAFT [60]2025.01 | 72.3 | |
| DSOptimized per sequence=true2022.05 | 72.1 | |
| DSInf. Input RGB=true, Inf. Input Motion=true, Input Resolution=240 × 426, Motion Est. Method=RAFT [60]2025.01 | 72.1 | |
| GWM*Post-process=CRF + SP‡2023.04 | 72 | |
| OCLRInf. RGB=true, Input Flow=true2022.05 | 71.6 | |
| OCLR+Post-process=DINO-based†2023.04 | 71.6 | |
| MODPost-process=DINO-based†2023.04 | 69.4 | |
| GWM*Post-process=None2023.04 | 69 | |
| OCLR+Post-process=None2023.04 | 67.6 | |
| OCLRInf. Input RGB=false, Inf. Input Motion=false, Input Resolution=112 × 224, Motion Est. Method=RAFT [60]2025.01 | 67.6 | |
| NLCFlow Method=SIFTFlow [48]2022.05 | 67.2 | |
| GWMPost-process=None2023.04 | 66.7 | |
| DivAInf. Input RGB=true, Inf. Input Motion=false, Input Resolution=128 × 224, Motion Est. Method=RAFT [60]2025.01 | 64.6 | |
| MODPost-process=None2023.04 | 62.2 | |
| CIS (post-processed)extra post-processing=true2022.05 | 62 | |
| SIMOPost-process=None2023.04 | 62 | |
| CISPost-process=CRF + SP‡2023.04 | 62 | |
| TokenCutPost-process=CRF only2023.04 | 61.6 | |
| TokenCutPost-process=None2023.04 | 59.6 | |
| MGInf. RGB=false, Input Flow=true2022.05 | 58.6 | |
| MGPost-process=None2023.04 | 58.6 | |
| SAGEFlow Method=LDOF [11]2022.05 | 57.6 | |
| SAGEPost-process=None2023.04 | 57.6 | |
| AMDInf. RGB=true, Input Flow=false2022.05 | 57 | |
| AMDPost-process=None2023.04 | 57 | |
| EMInf. RGB=false, Input Flow=true2022.05 | 55.5 | |
| Meunier et al.Inf. Input RGB=false, Inf. Input Motion=false, Input Resolution=128 × 224, Motion Est. Method=RAFT [60]2025.01 | 55 | |
| CUTFlow Method=LDOF [11]2022.05 | 54.3 | |
| CUTPost-process=None2023.04 | 54.3 | |
| CUTInf. Input RGB=true, Inf. Input Motion=true, Motion Est. Method=LDOF [5]2025.01 | 54.3 | |
| FTSFlow Method=LDOF [11]2022.05 | 47.8 | |
| FTSPost-process=None2023.04 | 47.8 | |
| FTSInf. Input RGB=true, Inf. Input Motion=true, Motion Est. Method=LDOF [5]2025.01 | 47.8 | |
| CISFlow Method=PWCNet [75]2022.05 | 45.6 | |
| CISPost-process=None2023.04 | 45.6 | |
| NLCe2e=false, Sup.=edge, Flow=SIFTFlow [67]2021.11 | 0.672 | |
| SAGEe2e=false, Flow=LDOF [66]2021.11 | 0.576 | |
| ARPe2e=false, Sup.=saliency, Flow=CPMFlow [69]2021.11 | 0.572 | |
| AMD (per-vid)e2e=true2021.11 | 0.57 | |
| CUTe2e=false, Flow=LDOF [66]2021.11 | 0.543 | |
| FTSe2e=false, Flow=LDOF [68]2021.11 | 0.478 | |
| CISe2e=false, Flow=PWC [20]2021.11 | 0.456 | |
| MGe2e=false, Flow=ARFlow [6]2021.11 | 0.378 | |
| AMD (per-img)e2e=false2021.11 | 0.287 |