Action Recognition on PKU-MMD II (xsub)
64.3AccuracyHeterogeneous Skeleton-Based Action Representation Learning
Evaluation results
| Method | Links | |
|---|---|---|
| Heterogeneous Skeleton-Based Action Representation LearningModality=J+C+S, Evaluation protocol=Transfer learning2025.06 | 64.3 | |
| Heterogeneous Skeleton-Based Action Representation LearningModality=J+C+S, Evaluation protocol=Transfer learning2025.06 | 63.1 | |
| 3s-USDRL (DSTE)Modality=Joint + Motion + Bone, Evaluation Protocol=Unsupervised2024.12 | 59.7 | |
| UmURLModality=J+M+B, Evaluation protocol=Transfer learning2025.06 | 59.7 | |
| UmURLModality=J+M+B, Evaluation protocol=Transfer learning2025.06 | 58.5 | |
| CMDModality=J, Evaluation protocol=Transfer learning2025.06 | 57 | |
| HiCoModality=J, Evaluation protocol=Transfer learning2025.06 | 56.3 | |
| CMDModality=J, Evaluation protocol=Transfer learning2025.06 | 56 | |
| 3s-SkeAttnCLRModality=J+M+B, FLOPs/(G)=21.96, Model Size/M=228.6, FPS=26.52025.12 | 55.5 | |
| HiCoModality=J, Evaluation protocol=Transfer learning2025.06 | 55.4 | |
| Decomposition and CompositionModality=J+M+B, FLOPs/(G)=2.54, Model Size/M=277.3, FPS=114.12025.12 | 54.7 | |
| USDRL (DSTE)Modality=Joint, Evaluation Protocol=Unsupervised2024.12 | 54.4 | |
| 3s-UmURLModality=Joint + Motion + Bone, Evaluation Protocol=Unsupervised2024.12 | 54.3 | |
| CrosSCLRModality=J, Evaluation protocol=Transfer learning2025.06 | 54 | |
| 3s-UmURLModality=J+M+B, FLOPs/(G)=2.54, Model Size/M=277.3, FPS=114.12025.12 | 54 | |
| MAMPModality=Joint, Evaluation Protocol=Unsupervised2024.12 | 53.8 | |
| 3s-HiCLRModality=Joint + Motion + Bone, Evaluation Protocol=Unsupervised2024.12 | 53.8 | |
| SkeAttnCLRModality=J, FLOPs/(G)=7.32, Model Size/M=76.2, FPS=79.52025.12 | 52.9 | |
| CrosSCLRModality=J, Evaluation protocol=Transfer learning2025.06 | 52.8 | |
| Decomposition and CompositionModality=J, FLOPs/(G)=1.74, Model Size/M=277.3, FPS=78.12025.12 | 52.7 | |
| UmURLModality=Joint + Motion + Bone, Evaluation Protocol=Unsupervised2024.12 | 52.6 | |
| 3s-CMDModality=Joint + Motion + Bone, Evaluation Protocol=Unsupervised2024.12 | 52.6 | |
| 3s-CMDModality=J+M+B, FLOPs/(G)=17.28, Model Size/M=946.5, FPS=7.42025.12 | 52.6 | |
| 3s-PSTLModality=J+M+B, FLOPs/(G)=3.45, Model Size/M=9.66, FPS=59.22025.12 | 52.3 | |
| UmURLModality=Joint, Evaluation Protocol=Unsupervised2024.12 | 52.1 | |
| UmURLModality=J, FLOPs/(G)=1.74, Model Size/M=277.3, FPS=78.12025.12 | 52.1 | |
| USDRL (STTR)Modality=Joint, Evaluation Protocol=Unsupervised2024.12 | 51.8 | |
| PCM3Modality=Joint, Evaluation Protocol=Unsupervised2024.12 | 51.5 | |
| 3s-CPMModality=J+M+B, FLOPs/(G)=6.662025.12 | 51.5 | |
| Masked ColorizationModality=Joint, Evaluation Protocol=Unsupervised2024.12 | 49.8 | |
| HiCoModality=Joint, Evaluation Protocol=Unsupervised2024.12 | 49.4 | |
| PSTLModality=J, FLOPs/(G)=1.15, Model Size/M=23.2, FPS=66.62025.12 | 49.3 | |
| CPMModality=J, FLOPs/(G)=2.222025.12 | 48.3 | |
| ISCModality=J, Evaluation protocol=Transfer learning2025.06 | 45.9 | |
| M2LModality=J, Evaluation protocol=Transfer learning2025.06 | 45.8 | |
| LongT GANModality=J, Evaluation protocol=Transfer learning2025.06 | 44.8 | |
| HaLPModality=Joint, Evaluation Protocol=Unsupervised2024.12 | 43.5 | |
| CMDModality=Joint, Evaluation Protocol=Unsupervised2024.12 | 43 | |
| CMDModality=J, FLOPs/(G)=5.76, Model Size/M=315.5, FPS=22.22025.12 | 43 | |
| 3s-AimCLRModality=J+M+B, FLOPs/(G)=3.45, Model Size/M=69.6, FPS=22.22025.12 | 39.5 | |
| AimCLRModality=Joint, Evaluation Protocol=Unsupervised2024.12 | 38.5 | |
| MS2LModality=Joint, Evaluation Protocol=Unsupervised2024.12 | 27.6 |