Pedestrian Detection on CityPersons (val)
17.9AP (Reasonable)InterNet
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| InterNetBackbone=ResNet-50, Input Scale=x12021.11 | 17.9 | 60.4 | 23.7 | 13.5 | — | — | — | 70.2 | 48.2 | |
| F.RCNN+ATT-vbbBackbone=VGG-16, Input Scale=x12021.11 | 16.4 | — | — | — | — | — | — | 57.3 | — | |
| F.RCNN+ATT-partBackbone=VGG-16, Input Scale=x12021.11 | 16 | — | — | — | — | — | — | 56.7 | 38.2 | |
| Faster RCNN2022.06 | 15.52 | — | — | — | — | — | — | 64.83 | 41.45 | |
| TLLModel Type=single-shot, MRF=false2018.07 | 15.5 | 53.6 | 17.2 | 10 | — | — | — | — | — | |
| TLLBackbone=ResNet-50, Input scale=1x2020.02 | 15.5 | 53.6 | 17.2 | 10 | — | — | — | — | — | |
| TLLBackbone=ResNet-50, Input Scale=x12021.11 | 15.5 | 53.6 | 17.2 | 10 | — | — | — | — | — | |
| TLL2022.10 | 15.5 | 53.6 | 17.2 | 10 | — | — | — | — | — | |
| TLLSize=x12022.10 | 15.5 | 53.6 | 17.2 | 10 | — | — | — | — | — | |
| Zhang et al.+Segmentation=true, Scale=x12017.11 | 15.4 | 55 | 18.9 | 9.3 | — | — | — | — | — | |
| Citypersons2018.07 | 15.4 | — | — | — | — | — | — | — | — | |
| FRCNNBackbone=VGG-16, Input scale=1x2020.02 | 15.4 | — | — | — | — | — | — | — | — | |
| Adapted FasterRCNNBackbone=VGG-16, Input Scale=x12021.11 | 15.4 | 55 | 18.9 | 9.3 | — | — | — | 64.8 | 41.5 | |
| Adapted Faster RCNNSize=x12022.10 | 15.4 | — | — | — | — | — | — | — | — | |
| Zhang et al.+Segmentation=false, Scale=x12017.11 | 14.8 | — | — | — | — | — | — | — | — | |
| FRCNN+SegBackbone=VGG-16, Input scale=1x2020.02 | 14.8 | — | — | — | — | — | — | — | — | |
| BaselineScale=x1, Backbone=ResNet-502017.11 | 14.6 | 60.6 | 18.6 | 7.9 | — | — | — | — | — | |
| TLL(MRF)Model Type=single-shot, MRF=true2018.07 | 14.4 | 52 | 15.9 | 9.2 | — | — | — | — | — | |
| TLL+MRFBackbone=ResNet-50, Input scale=1x2020.02 | 14.4 | 52 | 15.9 | 9.2 | — | — | — | — | — | |
| TLL+MRFBackbone=ResNet-50, Input Scale=x12021.11 | 14.4 | 52 | 15.9 | 9.2 | — | — | — | 52 | — | |
| Tao et al.2022.06 | 14.4 | — | — | — | — | — | — | 52 | 34.24 | |
| TLL + MRFSize=x12022.10 | 14.4 | 52 | 15.9 | 9.2 | — | — | — | — | — | |
| FC-Net2023.04 | 13.9 | — | — | — | — | — | — | — | 29.6 | |
| RepLoss+RepGT=true, Scale=x1, Backbone=ResNet-502017.11 | 13.7 | 57.5 | 17.3 | 7.2 | — | — | — | — | — | |
| RepLoss+RepBox=true, Scale=x1, Backbone=ResNet-502017.11 | 13.7 | 59.1 | 17.2 | 7.8 | — | — | — | — | — | |
| Das et al.2022.06 | 13.29 | — | — | — | — | — | — | 46.07 | 29.13 | |
| RepLoss+RepGT=true, +RepBox=true, Scale=x1, Backbone=ResNet-502017.11 | 13.2 | 56.9 | 16.8 | 7.6 | — | — | — | — | — | |
| Repulsion Loss2018.07 | 13.2 | 56.9 | 16.8 | 7.6 | — | — | — | — | — | |
| Repulsion LossBackbone=ResNet-50, Input scale=1x2020.02 | 13.2 | 56.9 | 16.8 | 7.6 | — | — | — | — | — | |
| Repulsion LossBackbone=ResNet-50, Input Scale=x12021.11 | 13.2 | 56.9 | 16.8 | 7.6 | — | — | — | 56.9 | — | |
| RepLoss2022.10 | 13.2 | 56.9 | 16.8 | 7.6 | — | — | — | — | — | |
| RepLossSize=x12022.10 | 13.2 | 56.9 | 16.8 | 7.6 | — | — | — | — | — | |
| Zhang et al.+Segmentation=false, Scale=x1.32017.11 | 12.8 | — | — | — | — | — | — | — | — | |
| OR-CNNBackbone=VGG-16, Input scale=1x2020.02 | 12.8 | 55.7 | 15.3 | 6.7 | — | — | — | — | — | |
| OR-CNNBackbone=VGG-16, Input Scale=x12021.11 | 12.8 | 55.7 | 15.3 | 6.7 | — | — | — | 55.7 | — | |
| OR-CNN2022.10 | 12.8 | 55.7 | 15.3 | 6.7 | — | — | — | — | — | |
| OR-CNNSize=x12022.10 | 12.8 | 55.7 | 15.3 | 6.7 | — | — | — | — | — | |
| Adapted Faster RCNNSize=x1.32022.10 | 12.8 | — | — | — | — | — | — | — | — | |
| HBANBackbone=VGG-16, Input scale=1x2020.02 | 12.5 | 48.1 | — | — | — | — | — | — | — | |
| CrowdDetBackbone=ResNet-50, Input Scale=x12021.11 | 12.1 | 47.7 | 12.9 | 7.5 | — | — | — | 40 | 25.4 | |
| Deep Multi-Task Network2022.06 | 12.01 | — | — | — | — | — | — | 44.7 | 27.8 | |
| ALFBackbone=ResNet-50, Input scale=1x2020.02 | 12 | 51.9 | 11.4 | 8.4 | — | — | — | — | — | |
| ALFNetBackbone=ResNet-50, Input Scale=x12021.11 | 12 | 51.9 | 11.4 | 8.4 | — | — | — | 43.8 | 26.3 | |
| ALFNet2022.10 | 12 | 51.9 | 11.4 | 8.4 | — | — | — | — | — | |
| Adaptive NMSSize=x12022.10 | 12 | 51.2 | 11.9 | 6.8 | — | — | — | — | — | |
| ALFNetSize=x12022.10 | 12 | 51.9 | 11.4 | 7.3 | — | — | — | — | — | |
| ALFNet2023.04 | 12 | — | — | — | — | — | — | 43.8 | 26.3 | |
| Adaptive NMSBackbone=ResNet-50, Input scale=1x2020.02 | 11.9 | 54 | 11.4 | 6.2 | — | — | — | — | — | |
| Adaptive-NMSBackbone=VGG-16, Input Scale=x12021.11 | 11.9 | 55.2 | 12.6 | 6.2 | — | — | — | 55.2 | — | |
| ClueNet2022.06 | 11.87 | — | — | — | — | — | — | 47.68 | 30.84 | |
| RepLoss+RepGT=true, +RepBox=true, Scale=x1.3, Backbone=ResNet-502017.11 | 11.6 | 55.3 | 14.8 | 7 | — | — | — | — | — | |
| RepLossSize=x1.32022.10 | 11.6 | 55.3 | 14.8 | 7 | — | — | — | — | — | |
| MGANBackbone=VGG-16, Input Scale=x12021.11 | 11.5 | — | — | — | — | — | — | 51.7 | — | |
| EGCLBackbone=VGG-16, Input Scale=x12021.11 | 11.5 | 51.1 | 11.9 | 6.1 | — | — | — | 50 | 27.9 | |
| HGPDBackbone=VGG-16, Input Scale=x12021.11 | 11.3 | — | — | — | — | — | — | 51.7 | — | |
| R2NMSBackbone=VGG-16, Input Scale=x12021.11 | 11.1 | — | — | — | — | — | — | 53.3 | — | |
| FRCN+A+DTBackbone=VGG-16, Input Scale=x1.32021.11 | 11.1 | 44.3 | 11.2 | 6.9 | — | — | — | — | — | |
| CSPBackbone=ResNet-50, Input scale=1x2020.02 | 11 | 49.3 | 10.4 | 7.3 | — | — | — | — | — | |
| CaseBackbone=VGG-16, Input Scale=x12021.11 | 11 | — | — | — | — | — | — | 50.3 | — | |
| OR-CNN2022.06 | 11 | — | — | — | — | — | — | 51 | 36.11 | |
| CSP2022.10 | 11 | 49.3 | 10.4 | 7.3 | — | — | — | — | — | |
| CSPSize=x12022.10 | 11 | 49.3 | 10.4 | 7.3 | — | — | — | — | — | |
| OR-CNNSize=x1.32022.10 | 11 | 51.3 | 13.7 | 5.9 | — | — | — | — | — | |
| RepLoss+RepGT=true, +RepBox=true, Scale=x1.5, Backbone=ResNet-502017.11 | 10.9 | 52.9 | 13.4 | 6.3 | — | — | — | — | — | |
| CrowdDet+EGCLBackbone=ResNet-50, Input Scale=x12021.11 | 10.9 | 46.4 | 11.6 | 7.4 | — | — | — | 39.3 | 24.8 | |
| EGCL2023.04 | 10.9 | — | — | — | — | — | — | 39.3 | 24.8 | |
| NOH-NMSBackbone=ResNet-50, Input Scale=x1.32021.11 | 10.8 | 53 | 11.2 | 6.6 | — | — | — | — | — | |
| Adaptive NMSSize=x1.32022.10 | 10.8 | 54 | 11.4 | 6.2 | — | — | — | — | — | |
| CrowdDetBackbone=ResNet-50, Input Scale=x1.32021.11 | 10.7 | 45.8 | 10.7 | 7.3 | — | — | — | 38 | 24.3 | |
| CrowdDetSize=x1.32022.10 | 10.7 | — | — | — | — | — | — | — | — | |
| PRNet++2023.04 | 10.7 | — | — | — | — | — | — | 40.9 | 25.4 | |
| APDBackbone=ResNet-50, Input scale=1x2020.02 | 10.6 | 49.8 | 9.5 | 7.1 | — | — | — | — | — | |
| Beta R-CNN2022.10 | 10.6 | 47.1 | 10.3 | 6.4 | — | — | — | — | — | |
| Beta R-CNNSize=x12022.10 | 10.6 | 47.1 | 10.3 | 6.4 | — | — | — | — | — | |
| MGANBackbone=VGG-16, Input scale=1x2020.02 | 10.5 | 47.2 | — | — | — | — | — | — | — | |
| CrowdDet+EGCLBackbone=ResNet-50, Input Scale=x1.32021.11 | 10.5 | 45.3 | 10.2 | 6.8 | — | — | — | 37.2 | 23.8 | |
| PSC-NetBackbone=VGG-16, Input scale=1x2020.02 | 10.4 | 39.7 | — | — | — | — | — | — | — | |
| MGANBackbone=VGG-16, Input Scale=x1.32021.11 | 10.3 | — | — | — | — | — | — | 49.6 | — | |
| ACSP (Smooth L1)Backbone=ResNet-101, Input scale=1x2020.02 | 10 | 46.1 | 8.8 | 6.7 | — | — | — | — | — | |
| Beta R-CNNSize=x1.32022.10 | 9.9 | 45.8 | 9.1 | 6 | — | — | — | — | — | |
| SMPD2023.04 | 9.9 | — | — | — | — | — | — | 36.6 | — | |
| CaseBackbone=VGG-16, Input Scale=x1.32021.11 | 9.6 | — | — | — | — | — | — | 48.2 | — | |
| VLPD2023.04 | 9.4 | — | — | — | — | — | — | 34.9 | 21.7 | |
| ACSP (Vanilla L1)Backbone=ResNet-101, Input scale=1x2020.02 | 9.3 | 46.3 | 8.7 | 5.6 | — | — | — | — | — | |
| APDBackbone=DLA-34, Input scale=1x2020.02 | 8.8 | 46.6 | 8.3 | 5.8 | — | — | — | — | — | |
| AdaptiveScale=x1.32021.04 | — | — | — | — | 10.8 | — | — | — | — | |
| Adaptive-NMSBackbone=VGG-16, Resolution=1.3x2020.03 | — | — | — | — | 10.8 | — | — | — | — | |
| AF-RCNNBackbone=VGG-16, Resolution=1.3x2020.03 | — | — | — | — | 12.8 | — | — | — | — | |
| AF-RCNNScale=x1.32021.04 | — | — | — | — | 12.81 | — | — | — | — | |
| ALFNetBackbone=Res-50, Scale=x12021.04 | — | — | — | — | 12 | — | — | — | — | |
| ATSS#Shot=Full2024.07 | — | — | — | — | 27.8 | 54.1 | — | — | — | |
| BaselineBackbone=Res-50, Scale=x1.32021.04 | — | — | — | — | 12.32 | 95.25 | — | — | — | |
| Bi-boxScale=x1.32021.04 | — | — | — | — | 11.24 | — | — | — | — | |
| Crowd-SAM#Shot=502024.07 | — | — | — | — | 31.7 | 33.3 | — | — | — | |
| CrowdDetBackbone=Res-50, Scale=x1.32021.04 | — | — | — | — | 10.7 | 96.1 | — | — | — | |
| D-DETRBackbone=Swin-L, #Queries=5002022.03 | — | — | — | — | 8.4 | 96.4 | 92 | — | — | |
| D-DETR + OursBackbone=Swin-L, #Queries=5002022.03 | — | — | — | — | 5.9 | 97.5 | 93.7 | — | — | |
| De-FRCN#Shot=502024.07 | — | — | — | — | 67.1 | 25.5 | — | — | — | |
| FCOS#Shot=Full2024.07 | — | — | — | — | 30 | 58.8 | — | — | — | |
| FPN + NMSBackbone=Res-50, Resolution=1.3x2020.03 | — | — | — | — | 11.7 | 95.2 | — | — | — |