Pedestrian Attribute Recognition on PA-100K
92.88mAVLM-PAR
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| VLM-PAR2025.12 | 92.88 | — | — | — | 92.32 | — | |
| PATHBackbone=ViT-L2023.12 | 90.8 | — | — | — | — | — | |
| PATHBackbone=ViT-L2023.12 | 90.8 | — | — | — | — | — | |
| PATHBackbone=ViT-L, Evaluation Protocol=Partial FT2023.03 | 90.8 | — | — | — | — | — | |
| AttriVision2025.12 | 89.8 | — | — | — | 93.1 | — | |
| Hulk-FTBackbone=ViT-L, Fine-tuned=true2023.12 | 88.97 | — | — | — | — | — | |
| Hulk-FTBackbone=ViT-L2023.12 | 88.97 | — | — | — | — | — | |
| UAPARPublish=Ours2026.04 | 88.48 | 82.83 | 87.97 | 93.09 | 90.46 | — | |
| KGPARPublish=arXiv252026.04 | 87.95 | 84.01 | 89.55 | 91.65 | 90.28 | — | |
| CLASP (RethinkPAR)Backbone=Swin-B2026.01 | 87.91 | — | — | — | — | — | |
| Hulk-FTBackbone=ViT-B, Fine-tuned=true2023.12 | 87.85 | — | — | — | — | — | |
| Hulk-FTBackbone=ViT-B2023.12 | 87.85 | — | — | — | — | — | |
| JGELPublish=AAAI262026.04 | 87.61 | 82.03 | — | 94.91 | 88.92 | — | |
| PromptPARPublish=TCSVT242026.04 | 87.47 | 83.78 | 89.27 | 91.7 | 90.15 | — | |
| CLEAR2025.12 | 87.2 | — | — | — | 91 | — | |
| PATHBackbone=ViT-B2023.12 | 86.9 | — | — | — | — | — | |
| PATHBackbone=ViT-B2023.12 | 86.9 | — | — | — | — | — | |
| PATHBackbone=ViT-B, Evaluation Protocol=Partial FT2023.03 | 86.9 | — | — | — | — | — | |
| CLASP (RethinkPAR)Backbone=Swin-S2026.01 | 86.88 | — | — | — | — | — | |
| HAPBackbone=ViT-B2023.12 | 86.54 | — | — | — | — | — | |
| HAPBackbone=ViT-B2023.12 | 86.54 | — | — | — | — | — | |
| HAPBackbone=ViT-B2026.01 | 86.54 | — | — | — | — | — | |
| CLIP+CP2025.12 | 86.45 | — | — | — | 92.77 | — | |
| SOLIDERBackbone=Swin-B2023.12 | 86.37 | — | — | — | — | — | |
| SOLIDERBackbone=Swin-B2023.12 | 86.37 | — | — | — | — | — | |
| SOLIDERBackbone=Swin-B2023.03 | 86.37 | — | — | — | — | — | |
| SOLIDERBackbone=Swin-B2026.01 | 86.37 | — | — | — | — | — | |
| SOLIDERBackbone=Swin-S2023.03 | 86.25 | — | — | — | — | — | |
| UniHCP-FTBackbone=ViT-B2023.12 | 86.18 | — | — | — | — | — | |
| UniHCP-FTBackbone=ViT-B2023.12 | 86.18 | — | — | — | — | — | |
| UniHCPBackbone=ViT-B2026.01 | 86.18 | — | — | — | — | — | |
| CLASP (RethinkPAR)Backbone=Swin-T2026.01 | 85.32 | — | — | — | — | — | |
| PATHBackbone=ViT-B, Evaluation Protocol=FT2023.03 | 85 | — | — | — | — | — | |
| PATHBackbone=ViT-B2026.01 | 85 | — | — | — | — | — | |
| HDFLPublish=NN252026.04 | 84.92 | 80.23 | 87.45 | 88.74 | 87.72 | — | |
| HulkBackbone=ViT-L, Fine-tuned=false2023.12 | 84.36 | — | — | — | — | — | |
| HulkBackbone=ViT-L2023.12 | 84.36 | — | — | — | — | — | |
| SOLIDERBackbone=Swin-T2023.03 | 84.14 | — | — | — | — | — | |
| FOCUSPublish=ICME252026.04 | 83.9 | 81.23 | 89.29 | 89.87 | 88.41 | — | |
| VTBPublish=TCSVT222026.04 | 83.72 | 80.89 | 87.88 | 89.3 | 88.21 | — | |
| ACMFNet2025.12 | 83.71 | — | — | — | 87.74 | — | |
| SSPNet2025.12 | 83.58 | — | — | — | 88.55 | — | |
| SSPNetPublish=PR242026.04 | 83.58 | 80.63 | — | 89.32 | 88.55 | — | |
| DAFLBackbone=ResNet-502023.12 | 83.54 | — | — | — | — | — | |
| DAFLBackbone=ResNet-502023.12 | 83.54 | — | — | — | — | — | |
| DAFLPublish=AAAI222026.04 | 83.54 | 80.13 | 87.01 | 89.19 | 88.09 | — | |
| SoTA2023.03 | 83.5 | — | — | — | — | — | |
| SOFAPublish=AAAI242026.04 | 83.4 | 81.1 | 88.4 | 89 | 88.3 | — | |
| AttExpIB-NetPublish=TIFS232026.04 | 83.23 | 79.42 | — | 88.6 | 87.23 | — | |
| HulkBackbone=ViT-B, Fine-tuned=false2023.12 | 82.85 | — | — | — | — | — | |
| HulkBackbone=ViT-B2023.12 | 82.85 | — | — | — | — | — | |
| HulkBackbone=ViT-B2026.01 | 82.85 | — | — | — | — | — | |
| APTMPre-training Dataset=MALS, MAM Loss Usage=Optimized2023.06 | 82.58 | 80.17 | 88.31 | 87.84 | 88.07 | — | |
| JLAC2023.03 | 82.31 | — | — | — | — | — | |
| MAEBackbone=ViT-B2023.03 | 82.3 | — | — | — | — | — | |
| L2LBackbone=ViT-B2023.12 | 82.24 | — | — | — | — | — | |
| L2LBackbone=ViT-B2023.12 | 82.24 | — | — | — | — | — | |
| IAAPublish=PR222026.04 | 81.94 | 80.31 | 88.36 | 88.01 | 87.8 | — | |
| SSCBackbone=ResNet-502023.12 | 81.87 | — | — | — | — | — | |
| SSCBackbone=ResNet-502023.12 | 81.87 | — | — | — | — | — | |
| RethinkPAR2023.03 | 81.61 | — | — | — | — | — | |
| RethinkPAREvaluation Protocol=Standard2023.06 | 81.61 | 79.45 | 87.66 | 87.59 | 87.62 | — | |
| RethinkPARBackbone=-2026.01 | 81.61 | — | — | — | — | — | |
| C-TranBackbone=ResNet-1012023.12 | 81.53 | — | — | — | — | — | |
| BaselinePre-training Dataset=Without pretraining on MALS2023.06 | 81.49 | 79.89 | 88.59 | 87.09 | 87.83 | — | |
| Q2LBackbone=TResNetL2023.12 | 80.72 | — | — | — | — | — | |
| ALMBackbone=BN-Inception2020.05 | 80.68 | 77.08 | 84.21 | 88.84 | 86.46 | — | |
| ALM2023.03 | 80.68 | — | — | — | — | — | |
| ALMEvaluation Protocol=Standard2023.06 | 80.68 | 77.08 | 84.21 | 88.84 | 86.46 | — | |
| EALCPublish=Neurocomputing232026.04 | 80.52 | 80.13 | — | 88.59 | 87.88 | — | |
| Baseline (wo MAM)MAM Loss Usage=MAM loss removed2023.06 | 80.43 | 79.91 | 89.26 | 86.49 | 87.85 | — | |
| MsVAA2023.03 | 80.41 | — | — | — | — | — | |
| MsVAABackbone=Res-1012026.01 | 80.41 | — | — | — | — | — | |
| BaselineBackbone=ResNet502020.05 | 79.38 | 78.56 | 89.41 | 84.78 | 86.55 | — | |
| strongBaselineEvaluation Protocol=Standard2023.06 | 79.38 | 78.56 | 89.41 | 84.78 | 86.55 | — | |
| UniHCPBackbone=ViT-B2023.12 | 79.32 | — | — | — | — | — | |
| UniHCPBackbone=ViT-B2023.12 | 79.32 | — | — | — | — | — | |
| VACBackbone=ResNet502020.05 | 79.16 | 79.44 | 88.97 | 86.26 | 87.59 | — | |
| VAC2023.03 | 79.16 | — | — | — | — | — | |
| VACBackbone=Res-1012026.01 | 79.16 | — | — | — | — | — | |
| FocalLossBackbone=ResNet502020.05 | 78.49 | 77.87 | 86.96 | 85 | 85.58 | — | |
| PATHBackbone=ViT-B, Evaluation Protocol=Head FT2023.03 | 77.4 | — | — | — | — | — | |
| LGNetBackbone=Inception-V22020.05 | 76.96 | 75.55 | 86.99 | 83.17 | 85.04 | — | |
| CLIPBackbone=ViT-B2023.03 | 76.1 | — | — | — | — | — | |
| PGDMBackbone=CaffeNet2020.05 | 74.95 | 73.08 | 84.36 | 82.24 | 83.29 | — | |
| HP-netModel type=HydraPlus-Net2017.09 | 74.21 | 72.19 | 82.97 | 82.09 | 82.53 | — | |
| HPNetBackbone=InceptionNet2020.05 | 74.21 | 72.19 | 82.97 | 82.09 | 82.53 | — | |
| HP-netEvaluation Protocol=Standard2023.06 | 74.21 | 72.19 | 82.97 | 82.09 | 82.53 | — | |
| DeepMarModel type=CNN2017.09 | 72.7 | 70.39 | 82.24 | 80.42 | 81.32 | — | |
| DeepMARBackbone=CaffeNet2020.05 | 72.7 | 70.39 | 82.24 | 80.42 | 81.32 | — | |
| M-netModel type=CNN2017.09 | 72.3 | 70.44 | 81.7 | 81.05 | 81.38 | — | |
| Baseline (Image only)Evaluation Protocol=Only finetuning the Image Encoder and its following MLPs2023.06 | 71.68 | 54.51 | 60.47 | 83.6 | 70.18 | — | |
| C-Tran2022.07 | — | 78.97 | — | — | 86.86 | — | |
| Label2Label2022.07 | — | 79.03 | — | — | 86.96 | — | |
| Q2L2022.07 | — | 78.78 | — | — | 86.73 | — | |
| SequencePARPublish=PR252026.04 | — | 83.94 | 90.38 | 90.23 | 90.1 | — | |
| ViT-REPublish=TMM242026.04 | — | 81.47 | 89.78 | 89.77 | 88.88 | — |