Action Recognition on PKU-MMD Part I
94.5Accuracy3s-ASMa
Evaluation results
| Method | Links | |
|---|---|---|
| 3s-ASMaBackbone=ST-GCN, Params=6.3, FLOPs=4.5, Evaluation Protocol=Lin.2026.02 | 94.5 | |
| STJD-CLBackbone=GCN, Evaluation Protocol=Lin.2026.02 | 93.2 | |
| MacDiffEvaluation Protocol=Linear Evaluation, Stream Ensemble=single-stream2026.03 | 92.8 | |
| 3s-S2IEvaluation Protocol=Linear Evaluation, Stream Ensemble=3-stream2026.03 | 92.3 | |
| MAMPInput stream=Joint only, Evaluation protocol=Linear2023.08 | 92.2 | |
| MAMPEvaluation Protocol=Linear Evaluation, Stream Ensemble=single-stream2026.03 | 92.2 | |
| SCD-NetBackbone=GCN+Transformer, Evaluation Protocol=Lin.2026.02 | 91.9 | |
| ActCLRBackbone=GCN, Evaluation Protocol=Lin.2026.02 | 91.6 | |
| 3s-CPMEvaluation Protocol=Linear Evaluation, Stream Ensemble=3-stream2026.03 | 90.7 | |
| AimCLR++Backbone=ST-GCN, Protocol=Linear Evaluation2026.03 | 90.4 | |
| ISSMBackbone=bi-GRU, Protocol=Linear Evaluation2026.03 | 89.6 | |
| 3s-PSTLBackbone=ST-GCN, Params=2.85, FLOPs=3.45, Evaluation Protocol=Lin.2026.02 | 89.2 | |
| M3GCLRBackbone=ST-GCN, Protocol=Linear Evaluation2026.03 | 89.1 | |
| CPMInput stream=Joint only, Evaluation protocol=Linear2023.08 | 88.8 | |
| CMCSBackbone=ST-GCN, Protocol=Linear Evaluation2026.03 | 88.1 | |
| S2IEvaluation Protocol=Linear Evaluation, Stream Ensemble=single-stream2026.03 | 88 | |
| 3s-AimCLRSupervision=Self-supervised, Evaluation protocol=Linear evaluation, Number of streams=3, Cross-stream knowledge mining=true2021.12 | 87.8 | |
| 3s-AimCLRBackbone=ST-GCN, Params=2.85, FLOPs=3.45, Evaluation Protocol=Lin.2026.02 | 87.8 | |
| AimCLRBackbone=ST-GCN, Protocol=Linear Evaluation2026.03 | 87.8 | |
| 3s-AimCLRSupervision=Self-supervised, Evaluation protocol=Linear evaluation, Number of streams=3, Cross-stream knowledge mining=false2021.12 | 87.4 | |
| 3s-AimCLRInput stream=Joint+Motion+Bone, Evaluation protocol=Linear2023.08 | 87.4 | |
| 3s-AimCLREvaluation Protocol=Linear Evaluation, Stream Ensemble=3-stream2026.03 | 87.4 | |
| 3s-AimCLRLabeled data percentage=10%, Evaluation protocol=Semi-supervised2021.12 | 86.1 | |
| Variational Contrastive LearningProtocol=Linear evaluation2026.01 | 86.1 | |
| 3s-SkeletonCLRInput stream=Joint+Motion+Bone, Evaluation protocol=Linear2023.08 | 85.3 | |
| 3s-SkeletonCLREvaluation Protocol=Linear Evaluation, Stream Ensemble=3-stream2026.03 | 85.3 | |
| 3s-CrosSCLRSupervision=Self-supervised, Evaluation protocol=Linear evaluation, Number of streams=3, Cross-stream knowledge mining=false2021.12 | 84.9 | |
| 3s-CrosSCLRInput stream=Joint+Motion+Bone, Evaluation protocol=Linear2023.08 | 84.9 | |
| 3s-CrosSCLRBackbone=ST-GCN, Params=26.6, FLOPs=17.28, Evaluation Protocol=Lin.2026.02 | 84.9 | |
| 3s-CrosSCLRProtocol=Linear evaluation2026.01 | 84.9 | |
| 3s-CrosSCLREvaluation Protocol=Linear Evaluation, Stream Ensemble=3-stream2026.03 | 84.9 | |
| CorsSCLRBackbone=ST-GCN, Protocol=Linear Evaluation2026.03 | 84.9 | |
| 3s-SkeletonBTBackbone=ST-GCN, Evaluation Protocol=Lin.2026.02 | 84.8 | |
| ST-GCNSupervision=Supervised, Evaluation protocol=Linear evaluation2021.12 | 84.1 | |
| VA-LSTMSupervision=Supervised, Evaluation protocol=Linear evaluation2021.12 | 84.1 | |
| SkeAttnCLRBackbone=ST-GCN, Protocol=Linear Evaluation2026.03 | 83.8 | |
| ASARBackbone=ST-GCN, Protocol=Linear Evaluation2026.03 | 83.5 | |
| Variational Contrastive LearningLabel percentage=10%2026.01 | 83.2 | |
| 3s-CrosSCLRLabeled data percentage=10%, Evaluation protocol=Semi-supervised2021.12 | 82.9 | |
| 3s-CrosSCLRLabel percentage=10%2026.01 | 82.9 | |
| SkeletonMAE*Input stream=Joint only, Evaluation protocol=Linear2023.08 | 82.8 | |
| SkeletonMAEEvaluation Protocol=Linear Evaluation, Stream Ensemble=single-stream2026.03 | 82.8 | |
| 3s-SkeletonCLRLabel percentage=10%2026.01 | 82.1 | |
| ISCSupervision=Self-supervised, Evaluation protocol=Linear evaluation2021.12 | 80.9 | |
| ISCInput stream=Joint only, Evaluation protocol=Linear2023.08 | 80.9 | |
| ISCProtocol=Linear evaluation2026.01 | 80.9 | |
| ISCEvaluation Protocol=Linear Evaluation, Stream Ensemble=single-stream2026.03 | 80.9 | |
| ISCLabeled data percentage=10%, Evaluation protocol=Semi-supervised2021.12 | 72.1 | |
| ISCLabel percentage=10%2026.01 | 72.1 | |
| MS2LLabeled data percentage=10%, Evaluation protocol=Semi-supervised2021.12 | 70.3 | |
| MS2LLabel percentage=10%2026.01 | 70.3 | |
| LongT GANLabeled data percentage=10%, Evaluation protocol=Semi-supervised2021.12 | 69.5 | |
| LongT GANLabel percentage=10%2026.01 | 69.5 | |
| LongT GANSupervision=Self-supervised, Evaluation protocol=Linear evaluation2021.12 | 67.7 | |
| LongT GANInput stream=Joint only, Evaluation protocol=Linear2023.08 | 67.7 | |
| LongT GANProtocol=Linear evaluation2026.01 | 67.7 | |
| LongT GANEvaluation Protocol=Linear Evaluation, Stream Ensemble=single-stream2026.03 | 67.7 | |
| MS2LSupervision=Self-supervised, Evaluation protocol=Linear evaluation2021.12 | 64.9 | |
| MS2LInput stream=Joint only, Evaluation protocol=Linear2023.08 | 64.9 | |
| MS2LProtocol=Linear evaluation2026.01 | 64.9 | |
| MS2LEvaluation Protocol=Linear Evaluation, Stream Ensemble=single-stream2026.03 | 64.9 | |
| P&CInput stream=Joint only, Evaluation protocol=Linear2023.08 | 59.9 | |
| P&CEvaluation Protocol=Linear Evaluation, Stream Ensemble=single-stream2026.03 | 59.9 | |
| 3s-AimCLRLabeled data percentage=1%, Evaluation protocol=Semi-supervised2021.12 | 57.5 | |
| Variational Contrastive LearningLabel percentage=1%2026.01 | 54.5 | |
| 3s-CrosSCLRLabeled data percentage=1%, Evaluation protocol=Semi-supervised2021.12 | 49.7 | |
| 3s-CrosSCLRLabel percentage=1%2026.01 | 49.7 | |
| ISCLabeled data percentage=1%, Evaluation protocol=Semi-supervised2021.12 | 37.7 | |
| ISCLabel percentage=1%2026.01 | 37.7 | |
| MS2LLabeled data percentage=1%, Evaluation protocol=Semi-supervised2021.12 | 36.4 | |
| MS2LLabel percentage=1%2026.01 | 36.4 | |
| LongT GANLabeled data percentage=1%, Evaluation protocol=Semi-supervised2021.12 | 35.8 | |
| LongT GANLabel percentage=1%2026.01 | 35.8 | |
| 3s-SkeletonCLRLabel percentage=1%2026.01 | 35.5 |