Video Object Segmentation on DAVIS 2016 (val)
91.8J MeanISVOS
Evaluation results
| Method | Links | ||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| ISVOSw/ BL30K=true2022.12 | 91.8 | — | 93.8 | — | — | — | — | — | 92.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ISVOSw/ BL30K=false2022.12 | 91.5 | — | 93.7 | — | — | — | — | — | 92.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ISVOSBL30K pre-training=false2024.09 | 91.5 | — | 93.7 | — | — | — | — | — | 92.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DropSegOnline Learning=false, Memory mechanism=false, Synthetic pre-training=false2023.04 | 90.9 | — | 93.3 | — | — | — | — | — | 92.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| STCNReference frames=Multiple, Resolution=480p, Training Data=DAVIS and YouTube-VOS2022.07 | 90.8 | — | 92.5 | — | — | — | — | — | 91.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| STCN2022.07 | 90.8 | — | 92.5 | — | — | — | — | — | 91.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| STCNPublication=NeurIPS 2021, Pre-training on image datasets=true, FPS=26*, Training with YouTube-VOS=true2022.08 | 90.8 | — | 92.5 | — | — | — | — | — | 91.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| STCNInit=mask, Online Learning=false, Memory Bank=true2022.07 | 90.8 | — | 92.5 | — | — | — | — | — | 91.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| STCNw/ BL30K=false2022.12 | 90.8 | — | 92.5 | — | — | — | — | — | 91.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| STCNLabeled data=100%2023.03 | 90.8 | — | 92.5 | — | — | — | — | — | 91.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| STCNFPS=27.2, Resolution=600p2022.03 | 90.8 | — | 92.5 | — | — | — | — | — | 91.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SwinB-AOT-LTraining on YouTube-VOS=true, Backbone=SwinB, Model Variant=Large, FPS=12.12021.06 | 90.7 | — | 93.3 | — | — | — | — | — | 92 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| QDMNsynthetic training dataset=true2022.07 | 90.7 | — | 93.2 | — | — | — | — | — | 92 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| XMemw/ BL30K=true2022.12 | 90.7 | — | 93.2 | — | — | — | — | — | 92 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BATMANCC=false2023.03 | 90.7 | — | — | — | — | — | — | — | 92.5 | — | — | — | — | — | — | — | — | 94.2 | — | — | — | — | |
| XMemCC=false, YouTube-VOS training=true, BL30K training=true2023.03 | 90.7 | — | — | — | — | — | — | — | 92 | — | — | — | — | — | — | — | — | 93.2 | — | 29.6 | — | — | |
| STCN*Labeled data=100%2023.03 | 90.7 | — | 92.3 | — | — | — | — | — | 91.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AOTTraining Data=with video data, Venue=NeurIPS'212023.05 | 90.7 | — | 93.3 | — | — | — | — | — | 92 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| XMemTraining Data=with video data, Venue=ECCV'222023.05 | 90.7 | — | 93.2 | — | — | — | — | — | 92 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| 2-shot STCN w/ OursLabeled data=2.9%2023.03 | 90.6 | — | 92 | — | — | — | — | — | 91.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AOSTFPS=17.5, Backbone=ResNet-50, Number of blocks=32022.03 | 90.6 | — | 93.6 | — | — | — | — | — | 92.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AOSTFPS=12.0, Backbone=Swin-B, Number of blocks=3, Parameter Sharing=true2022.03 | 90.6 | — | 94.1 | — | — | — | — | — | 92.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AOST-LBackbone=ResNet502024.09 | 90.6 | — | 93.6 | — | — | — | — | — | 92.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| JIMDBackbone=ResNet502024.09 | 90.6 | — | 94.2 | — | — | — | — | — | 92.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SwiftNet2022.07 | 90.5 | — | 90.3 | — | — | — | — | — | 90.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SwiftPublication=CVPR 2021, Pre-training on image datasets=true, FPS=25, Training with YouTube-VOS=true2022.08 | 90.5 | — | 90.3 | — | — | — | — | — | 90.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeAOTw/ BL30K=false2022.12 | 90.5 | — | 94 | — | — | — | — | — | 92.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SwiftNetCC=true, YouTube-VOS training=true2023.03 | 90.5 | — | — | — | — | — | — | — | 90.4 | — | — | — | — | — | — | — | — | 90.3 | — | 25 | — | — | |
| AOSTFPS=12.0, Backbone=Swin-B, Number of blocks=32022.03 | 90.5 | — | 94.2 | — | — | — | — | — | 92.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeAOT-LBackbone=ResNet502024.09 | 90.5 | — | 94 | — | — | — | — | — | 92.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| XMemw/ BL30K=false2022.12 | 90.4 | — | 92.7 | — | — | — | — | — | 91.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| STCNw/ BL30K=true2022.12 | 90.4 | — | 93 | — | — | — | — | — | 91.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| STCNCC=false, YouTube-VOS training=true, BL30K training=true2023.03 | 90.4 | — | — | — | — | — | — | — | 91.7 | — | — | — | — | — | — | — | — | 93 | — | 26.9 | — | — | |
| XMemCC=false, YouTube-VOS training=true2023.03 | 90.4 | — | — | — | — | — | — | — | 91.5 | — | — | — | — | — | — | — | — | 92.7 | — | 29.6 | — | — | |
| XMemLabeled data=100%2023.03 | 90.4 | — | 92.7 | — | — | — | — | — | 91.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| STCN2024.09 | 90.4 | — | 93 | — | — | — | — | — | 91.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| XMemBL30K pre-training=false2024.09 | 90.4 | — | 92.7 | — | — | — | — | — | 91.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MobileVOS (ResNet18*)CC=true, Backbone=ResNet-18, YouTube-VOS training=true, BL30K training=true2023.03 | 90.3 | — | — | — | — | — | — | — | 91.4 | — | — | — | — | — | — | — | — | 92.6 | — | 100.1 | — | — | |
| XMem*Labeled data=100%2023.03 | 90.3 | — | 92.4 | — | — | — | — | — | 91.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| 2-shot XMem w/ OursLabeled data=2.9%2023.03 | 90.3 | — | 92.3 | — | — | — | — | — | 91.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AOSTFPS=17.5, Backbone=ResNet-50, Number of blocks=3, Parameter Sharing=true2022.03 | 90.3 | — | 92.5 | — | — | — | — | — | 91.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| QDMNsynthetic training dataset=false2022.07 | 90.2 | — | 91.7 | — | — | — | — | — | 91 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MobileVOS (model soup)CC=true, Configuration=model soup, YouTube-VOS training=true2023.03 | 90.2 | — | — | — | — | — | — | — | 91.3 | — | — | — | — | — | — | — | — | 92.5 | — | 100.1 | — | — | |
| R50-AOT-LTraining on YouTube-VOS=true, Backbone=ResNet-50, Model Variant=Large, FPS=18.02021.06 | 90.1 | — | 92.1 | — | — | — | — | — | 91.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AOTw/ BL30K=false2022.12 | 90.1 | — | 92.1 | — | — | — | — | — | 91.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GSFMCC=false2023.03 | 90.1 | — | — | — | — | — | — | — | 91.4 | — | — | — | — | — | — | — | — | 92.7 | — | 8.9 | — | — | |
| R50-AOT-LLabeled data=100%2023.03 | 90.1 | — | 92.1 | — | — | — | — | — | 91.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GSFMFPS=8.92022.03 | 90.1 | — | 92.7 | — | — | — | — | — | 91.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AOSTFPS=18.2, Number of blocks=32022.03 | 90.1 | — | 93 | — | — | — | — | — | 91.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AOT-LBackbone=ResNet502024.09 | 90.1 | — | 92.1 | — | — | — | — | — | 91.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RDEw/ BL30K=true2022.12 | 90 | — | 93.2 | — | — | — | — | — | 91.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RDE-VOSCC=true, YouTube-VOS training=true, BL30K training=true2023.03 | 90 | — | — | — | — | — | — | — | 91.6 | — | — | — | — | — | — | — | — | 93.2 | — | 35 | — | — | |
| 2-shot RDE-VOS w/ OursLabeled data=2.9%2023.03 | 90 | — | 92 | — | — | — | — | — | 90.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LCMTime(s)=0.1182021.04 | 89.9 | — | 91.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 90.7 | — | |
| LCMResolution=480p, External Training Datasets=false2021.12 | 89.9 | — | 91.4 | — | — | — | — | — | 90.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LCMReference frames=Multiple, Resolution=480p, Training Data=DAVIS and YouTube-VOS2022.07 | 89.9 | — | 91.4 | — | — | — | — | — | 90.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LCM2022.07 | 89.9 | — | 91.4 | — | — | — | — | — | 90.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LCMPublication=CVPR 2021, Pre-training on image datasets=true, FPS=9, Training with YouTube-VOS=true2022.08 | 89.9 | — | 91.4 | — | — | — | — | — | 90.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SWEMPublication=CVPR 2022, Pre-training on image datasets=true, FPS=36, Training with YouTube-VOS=true2022.08 | 89.9 | — | 92.6 | — | — | — | — | — | 91.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SWEMTraining Data=with video data, Venue=CVPR'222023.05 | 89.9 | — | 92.6 | — | — | — | — | — | 91.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MiVOSResolution=480p, External Training Datasets=true2021.12 | 89.7 | — | 92.1 | — | — | — | — | — | 91 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MiVOSReference frames=Multiple, Resolution=480p, Training Data=DAVIS and YouTube-VOS2022.07 | 89.7 | — | 92.1 | — | — | — | — | — | 91 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MiVOSsynthetic training dataset=true2022.07 | 89.7 | — | 92.4 | — | — | — | — | — | 91 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AOT-L2022.07 | 89.7 | — | 92.3 | — | — | — | — | — | 91 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AOTPublication=NeurIPS 2021, Pre-training on image datasets=true, FPS=19, Training with YouTube-VOS=true2022.08 | 89.7 | — | 92.3 | — | — | — | — | — | 91 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RDEw/ BL30K=false2022.12 | 89.7 | — | 92.5 | — | — | — | — | — | 91.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MiVOSCC=false, YouTube-VOS training=true, BL30K training=true2023.03 | 89.7 | — | — | — | — | — | — | — | 91 | — | — | — | — | — | — | — | — | 92.4 | — | 16.9 | — | — | |
| RDE-VOSCC=true, YouTube-VOS training=true2023.03 | 89.7 | — | — | — | — | — | — | — | 91.1 | — | — | — | — | — | — | — | — | 92.5 | — | 35 | — | — | |
| MobileVOS (ResNet18)CC=true, Backbone=ResNet-18, YouTube-VOS training=true2023.03 | 89.7 | — | — | — | — | — | — | — | 90.6 | — | — | — | — | — | — | — | — | 91.6 | — | 100.1 | — | — | |
| RDE-VOSLabeled data=100%2023.03 | 89.7 | — | 92.5 | — | — | — | — | — | 91.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CFBIYouTube-VOS training=true, Multi-scale and flip strategy=true, Inference time (t/s)=92020.03 | 89.6 | 90.7 | 91.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AOT-LargeTraining on YouTube-VOS=true, Model Variant=Large, FPS=18.72021.06 | 89.6 | — | 91.1 | — | — | — | — | — | 90.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HMMN2022.07 | 89.6 | — | 92 | — | — | — | — | — | 90.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HMMNPublication=ICCV 2021, Pre-training on image datasets=true, FPS=10, Training with YouTube-VOS=true2022.08 | 89.6 | — | 92 | — | — | — | — | — | 90.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HMMNInit=mask, Online Learning=false, Memory Bank=true2022.07 | 89.6 | — | 92 | — | — | — | — | — | 90.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HMMNw/ BL30K=false2022.12 | 89.6 | — | 92 | — | — | — | — | — | 90.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MIVOSw/ BL30K=true2022.12 | 89.6 | — | 92.4 | — | — | — | — | — | 91 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MiVOSLabeled data=100%2023.03 | 89.6 | — | 92.4 | — | — | — | — | — | 91 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HMMNLabeled data=100%2023.03 | 89.6 | — | 92 | — | — | — | — | — | 90.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HMMNFPS=10.02022.03 | 89.6 | — | 92 | — | — | — | — | — | 90.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AOT-LFPS=18.72022.03 | 89.6 | — | 91.1 | — | — | — | — | — | 90.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HMMN2024.09 | 89.6 | — | 92 | — | — | — | — | — | 90.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| KMNTraining Data=+Youtube-VOS, OL=false, Time=0.12s2020.07 | 89.5 | — | 91.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 90.5 | — | — | — | |
| KMNTime(s)=0.122021.04 | 89.5 | — | 91.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 90.5 | — | |
| KMNYouTube-VOS training=true2021.03 | 89.5 | — | 91.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 90.5 | |
| KMNVOSSimulated data (S)=true, YouTube-VOS training (Y)=true, Inference Time (t/s)=0.122020.10 | 89.5 | 90.5 | 91.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| KMNTraining on YouTube-VOS=true, FPS=8.32021.06 | 89.5 | — | 91.5 | — | — | — | — | — | 90.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| KMN2022.07 | 89.5 | — | 91.5 | — | — | — | — | — | 90.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| KMNPublication=ECCV 2020, Pre-training on image datasets=true, FPS=8, Training with YouTube-VOS=true2022.08 | 89.5 | — | 91.5 | — | — | — | — | — | 90.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MobileVOS (MobileNetV2)CC=true, Backbone=MobileNetV2, YouTube-VOS training=true2023.03 | 89.5 | — | — | — | — | — | — | — | 90.5 | — | — | — | — | — | — | — | — | 91.5 | — | 81.8 | — | — | |
| RDE-VOS*Labeled data=100%2023.03 | 89.5 | — | 92.4 | — | — | — | — | — | 91 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| KMN2024.09 | 89.5 | — | 91.5 | — | — | — | — | — | 90.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MobileVOS (wo/ ASPP)CC=true, Configuration=without ASPP, YouTube-VOS training=true2023.03 | 89 | — | — | — | — | — | — | — | 90.1 | — | — | — | — | — | — | — | — | 91.1 | — | 86 | — | — | |
| RMNetYouTube-VOS training=true2021.03 | 88.9 | — | 88.7 | — | — | — | — | — | — | — | — | — | — | — | — | 0.084 | — | — | — | — | — | 88.8 | |
| RMNResolution=480p, External Training Datasets=true2021.12 | 88.9 | — | 88.7 | — | — | — | — | — | 88.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RMNetYouTube-VOS Training (Y)=true, Resolution=480p, Fine-tuning (Ft)=false, Simulated data (S)=true2021.12 | 88.9 | — | 88.7 | — | — | — | — | — | 88.8 | — | — | — | — | — | — | 0.084 | — | — | — | — | — | — | |
| RMNReference frames=Multiple, Resolution=480p, Training Data=DAVIS and YouTube-VOS2022.07 | 88.9 | — | 88.7 | — | — | — | — | — | 88.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RMNet2022.07 | 88.9 | — | 88.7 | — | — | — | — | — | 88.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RMNetPublication=CVPR 2021, Pre-training on image datasets=true, FPS=12, Training with YouTube-VOS=true2022.08 | 88.9 | — | 88.7 | — | — | — | — | — | 88.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RMNetCC=false, YouTube-VOS training=true2023.03 | 88.9 | — | — | — | — | — | — | — | 88.8 | — | — | — | — | — | — | — | — | 88.7 | — | 11.9 | — | — |