Video Object Detection on ImageNet VID (val)
91.3mAP (%)Location Anticipation
Evaluation results
| Method | Links | ||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Location AnticipationBackbone=SwinB, No Post-proc.=true, Detector=Deformable DETR2023.08 | 91.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 18.1 | — | — | |
| TGBFormerBackbone=Swin-B, Base Detector=DETR2025.03 | 90.3 | 49.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TransVOD LiteBackbone=SwinB, Runtime (FPS)=14.9, #Params (M)=106.32022.01 | 90.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TransVODBackbone=SwinB, No Post-proc.=true, Detector=Deformable DETR2023.08 | 90.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 14.9 | — | — | |
| TransVOD LiteBackbone=Swin-B, Base Detector=Deformable DETR2025.03 | 90.1 | 67.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TransVOD++Base Detector=Deformable DETR, Backbone=Swin-Base2022.01 | 90 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTSEFormerBase Detector=Deformable DETR, Stages=1, Backbone=ResNet-1012022.09 | 88.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTSEFormerBase Detector=Deformable DETR, Stages=1, Backbone=ResNet-101, Post-processing=None2022.09 | 88.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTSEFormerBackbone=R101, No Post-proc.=false, Detector=Deformable DETR2023.08 | 88.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Location AnticipationBackbone=R101, No Post-proc.=true, Detector=Deformable DETR2023.08 | 87.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 36.4 | — | — | |
| PTSEFormerBase Detector=Deformable DETR, Stages=1, Backbone=ResNet-502022.09 | 87.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Location AnticipationBackbone=R101, No Post-proc.=true, Detector=Faster RCNN2023.08 | 87.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | 0.78 | 39.6 | — | — | |
| TGBFormerBackbone=ResNet-101, Base Detector=DETR2025.03 | 86.5 | 24.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TransVOD LiteBackbone=SwinS, Runtime (FPS)=22.2, #Params (M)=68.32022.01 | 85.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ClipVIDICA module=Oracle2023.08 | 85.8 | — | — | 90.8 | 84.5 | 72.3 | — | — | — | — | — | — | — | — | — | — | — | — | |
| ClipVIDBackbone=X-101, fps=25.12023.08 | 85.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HVRNet + SeqNMSBackbone=X-101, Sequence-level post-processing=SeqNMS, fps=~ 12023.08 | 85.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MEGABackbone=ResNeXt-1012020.03 | 85.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MEGA + BLRBackbone=X-101, Sequence-level post-processing=BLR, fps=~ 12023.08 | 85.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CDANetBackbone=ResNet-101, Base Detector=Faster R-CNN2025.03 | 85.4 | 80.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MEGABackbone=X101, Post-processing=Seq-NMS2021.09 | 85.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CSMNBackbone=ResNet-101, Base Detector=Faster R-CNN2025.03 | 85.2 | 909.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CFA-NetBase Detector=Faster RCNN, Backbone=ResNet-1012022.01 | 85 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CFANetBackbone=ResNet-101, Base Detector=Faster R-CNN2025.03 | 85 | 884.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DGRNetBackbone=ResNet-101, Base Detector=Faster R-CNN2025.03 | 85 | 91.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HVRNetBase Detector=Faster-RCNN, Stages=2, Backbone=ResNet-101, Post-processing=Seq-NMS2022.09 | 84.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TROI + BoxMaskBackbone=ResNeXt-101, Base Detector=Faster R-CNN2022.10 | 84.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EBFABase Detector=Faster RCNN, Backbone=ResNet-1012022.01 | 84.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HVRNetBackbone=X-1012023.08 | 84.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EBFABackbone=ResNet-101, Base Detector=Faster R-CNN2025.03 | 84.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RDNBackbone=ResNeXt-1012020.03 | 84.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ClipVIDICA module=True2023.08 | 84.7 | — | — | 89.9 | 83.9 | 68.5 | — | — | — | — | — | — | — | — | — | — | — | — | |
| ClipVIDBackbone=R-101, fps=39.32023.08 | 84.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RDN + BLRBackbone=X-101, Sequence-level post-processing=BLR, fps=~ 12023.08 | 84.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ClipVIDBackbone=ResNet-101, Base Detector=DETR2025.03 | 84.7 | 25.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAMBABase Detector=Faster RCNN, Backbone=ResNet-1012022.01 | 84.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAMBABackbone=ResNet-101, Base Detector=Faster R-CNN2025.03 | 84.6 | 110.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MEGABackbone=ResNet-1012020.03 | 84.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MEGABase Detector=Faster-RCNN, Stages=2, Backbone=ResNet-101, Post-processing=Seq-NMS2022.09 | 84.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MEGABackbone=R101, No Post-proc.=false2023.08 | 84.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | 6.34 | 5.3 | — | — | |
| MEGA + BLRBackbone=R-101, Sequence-level post-processing=BLR, fps=~ 12023.08 | 84.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SELSABackbone=ResNeXt-101, Data Augmentation=true2019.07 | 84.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TROIBackbone=ResNeXt-101, Base Detector=Faster R-CNN2022.10 | 84.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SELSABackbone=X-1012023.08 | 84.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SELSA + TROIBackbone=X1012021.09 | 84.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MEGABackbone=ResNeXt-101, local=true, global=true2020.03 | 84.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DSFNetBase Detector=Faster RCNN, Backbone=ResNet-1012022.01 | 84.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DSFNet2023.08 | 84.1 | — | — | 90 | 82.6 | 67 | — | — | — | — | — | — | — | — | — | — | — | — | |
| DSFNetBackbone=R-101, fps=< 5.82023.08 | 84.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MEGABackbone=X-101, fps=5.32023.08 | 84.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DSFNetBackbone=ResNet-101, Base Detector=Faster R-CNN2025.03 | 84.1 | 384.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LeveragingBackbone=X101-FPN2021.09 | 84.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MEGABackbone=X1012021.09 | 84.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RDNBackbone=ResNet-1012020.03 | 83.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TF-BlenderBase Detector=Faster RCNN, Backbone=ResNet-1012022.01 | 83.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RDNBackbone=R101, No Post-proc.=false2023.08 | 83.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HVRNetBackbone=R101, No Post-proc.=false2023.08 | 83.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TF-BlenderBackbone=R101, No Post-proc.=false2023.08 | 83.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 4.9 | — | — | |
| TF-BlenderBackbone=R-101, fps=5.82023.08 | 83.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RDN + BLRBackbone=R-101, Sequence-level post-processing=BLR, fps=~ 12023.08 | 83.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HVRNet + SeqNMSBackbone=R-101, Sequence-level post-processing=SeqNMS, fps=~ 12023.08 | 83.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TransVOD LiteBackbone=SwinT, Runtime (FPS)=29.6, #Params (M)=46.92022.01 | 83.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Integrated detection and trackinginference=on-line, backbone=ResNet-1012018.11 | 83.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SELSA + BoxMaskBackbone=ResNeXt-101, Base Detector=Faster R-CNN2022.10 | 83.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ClipVIDICA module=False2023.08 | 83.3 | — | — | 89 | 82.7 | 66.1 | — | — | — | — | — | — | — | — | — | — | — | — | |
| MSTFBackbone=ResNet-101, Base Detector=Faster R-CNN2025.03 | 83.3 | 105.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RDNBackbone=ResNeXt-101, local=true, global=false2020.03 | 83.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HVRNetBase Detector=Faster-RCNN, Stages=2, Backbone=ResNet-1012022.09 | 83.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TROI + BoxMaskBackbone=ResNet-101, Base Detector=Faster R-CNN2022.10 | 83.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HVRNetBase Detector=Faster RCNN, Backbone=ResNet-1012022.01 | 83.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BoxMaskBackbone=R101, No Post-proc.=false2023.08 | 83.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 6.1 | — | — | |
| HVRNet2023.08 | 83.2 | — | — | 88.7 | 82.3 | 66.6 | — | — | — | — | — | — | — | — | — | — | — | — | |
| HVRNetBackbone=R-1012023.08 | 83.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RDNBackbone=X-1012023.08 | 83.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HVRNetBackbone=ResNet-101, Base Detector=Faster R-CNN2025.03 | 83.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RDNBackbone=X1012021.09 | 83.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SELSABackbone=ResNeXt-1012019.07 | 83.11 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SELSABackbone=X1012021.09 | 83.11 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| NUS-Qihoo-UIUCinference=off-line, backbone=DPN-1312018.11 | 83.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SELSABackbone=ResNeXt-101, Base Detector=Faster R-CNN2022.10 | 83.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MEGABackbone=ResNet-101, local=true, global=true2020.03 | 82.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MEGABase Detector=Faster-RCNN, Stages=2, Backbone=ResNet-1012022.09 | 82.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MEGABase Detector=Faster RCNN, Backbone=ResNet-1012022.01 | 82.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MEGA2023.08 | 82.9 | — | — | 89.4 | 81.6 | 62.7 | — | — | — | — | — | — | — | — | — | — | — | — | |
| MEGABackbone=R-101, fps=5.82023.08 | 82.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MEGABackbone=ResNet-101, Base Detector=Faster R-CNN2025.03 | 82.9 | 114.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MEGABackbone=R1012021.09 | 82.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SELSA2023.08 | 82.7 | — | — | 88 | 81.4 | 67.1 | — | — | — | — | — | — | — | — | — | — | — | — | |
| SELSABackbone=R-101, fps=9.62023.08 | 82.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SELSABackbone=ResNet-101, Data Augmentation=true2019.07 | 82.69 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SELSABase Detector=Faster-RCNN, Stages=2, Backbone=ResNet-101, Post-processing=Seq-NMS2022.09 | 82.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TSFABackbone=ResNet-101, Base Detector=Faster R-CNN2025.03 | 82.5 | 125 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MEGA + BoxMaskBackbone=ResNet-101, Base Detector=Faster R-CNN2022.10 | 82.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| QueryPropBackbone=ResNet-101, Base Detector=Sparse R-CNN2025.03 | 82.3 | 37.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Single-frame Detector + TROIBackbone=X1012021.09 | 82.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| D&TBackbone=Inception-v4, Post-processing=tubelet rescoring2019.07 | 82.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| D&TBackbone=Inception-v42020.03 | 82 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TROI + SELSABase Detector=Faster-RCNN, Stages=2, Backbone=ResNet-1012022.09 | 82 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TROIBackbone=ResNet-101, Base Detector=Faster R-CNN2022.10 | 82 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TROIBase Detector=Faster RCNN, Backbone=ResNet-1012022.01 | 82 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — |