Skeleton Based Action Recognition On Ntu Rgbd 1
評価指標
Accuracy (Cross-Setup)
Accuracy (Cross-Subject)
GFLOPS per prediction
評価結果
このベンチマークにおける各モデルのパフォーマンス結果
比較表
| モデル名 | Accuracy (Cross-Setup) | Accuracy (Cross-Subject) | GFLOPS per prediction |
|---|---|---|---|
| online-skeleton-based-action-recognition-with | 81.8 | 80.2 | 16.2 |
| jointly-learning-heterogeneous-features-for-1 | 54.7% | 50.8% | - |
| dg-stgcn-dynamic-spatial-temporal-modeling | 91.3 | 89.6 | - |
| joint-mixing-data-augmentation-for-skeleton | 91.9 | 90.9 | - |
| online-skeleton-based-action-recognition-with | - | 79 | 16.73 |
| hyperbolic-self-paced-learning-for-self | 86.3 | 84.5 | - |
| online-skeleton-based-action-recognition-with | 80.7 | 79.7 | 18.69 |
| graph-contrastive-learning-for-skeleton-based | 91.0 | 89.5 | - |
| constructing-stronger-and-faster-baselines | 84.3 | 85.9 | - |
| skeleton-image-representation-for-3d-action | 62.8% | 67.9% | - |
| skeleton-based-human-action-recognition-with | 63.3% | 61.2% | - |
| constructing-stronger-and-faster-baselines | 89.1 | 88.7 | - |
| spatio-temporal-lstm-with-trust-gates-for-3d | 57.9% | 55.7% | - |
| vertex-feature-encoding-and-hierarchical | 79.8% | 78.3% | - |
| a-new-representation-of-skeleton-sequences | 57.9% | 58.4% | - |
| skeleton-based-action-recognition-via | 90.8 | 89.4 | - |
| learning-multi-granular-spatio-temporal-graph | 89.3 | 88.2 | - |
| skateformer-skeletal-temporal-transformer-for | 91.4 | 89.8 | - |
| infogcn-representation-learning-for-human | 91.2 | 89.8 | - |
| gimme-signals-discriminative-signal-encoding | 71.6% | 70.8% | - |
| skeleton-based-action-recognition-using | 60.9% | 58.2% | - |
| feedback-graph-convolutional-network-for | 87.4% | 85.4% | - |
| early-action-prediction-by-soft-regression | 44.9% | 36.3% | - |
| recognizing-human-actions-as-the-evolution-of | 66.9% | 64.6% | - |
| channel-wise-topology-refinement-graph | 90.6 | 88.9 | - |
| degcn-deformable-graph-convolutional-networks | 92.1 | 91.0 | - |
| multi-scale-spatial-temporal-convolutional | 88.3 | 87.4 | - |
| shap-mix-shapley-value-guided-mixing-for-long | 91.7 | 90.4 | - |
| enhanced-skeleton-visualization-for-view | 63.2% | 60.3% | - |
| skeleton-based-action-recognition-with-shift | 87.6% | 85.9% | - |
| revisiting-skeleton-based-action-recognition | 90.3 | 86.9 | - |
| pyskl-towards-good-practices-for-skeleton | 90.8 | 88.6 | - |
| usdrl-unified-skeleton-based-dense | 79.3 | 80.6 | - |
| ntu-rgbd-a-large-scale-dataset-for-3d-human | 26.3% | 25.5% | - |
| revealing-key-details-to-see-differences-a | 92.2 | 90.9 | - |
| spatial-temporal-graph-convolutional-networks-1 | 88.4 | 86.2 | - |
| language-knowledge-assisted-representation | 91.8 | 90.7 | - |
| online-skeleton-based-action-recognition-with | 86.1 | 84.8 | 0.3 |
| spatial-temporal-graph-attention-network-for | 90.4 | 88.7 | - |
| online-skeleton-based-action-recognition-with | 81.6 | 79.4 | 0.16 |
| action-recognition-with-multi-stream-motion | 91.0 | 89.7 | - |
| learning-clip-representations-for-skeleton | 61.8% | 62.2% | - |
| leveraging-third-order-features-in-skeleton | 89.2% | 88.2% | - |
| skelemotion-a-new-representation-of-skeleton | 66.9% | 67.7% | - |
| online-skeleton-based-action-recognition-with | 85.5 | 84.0 | 0.32 |
| skeleton-image-representation-for-3d-action | 59.7% | 65.5% | - |
| step-catformer-spatial-temporal-effective | 91.2 | 90.0 | - |
| spatial-temporal-transformer-network-for | 84.7% | 82.7% | - |
| learning-stochastic-differential-equations-1 | 67.2% | 68.3% | - |
| disentangling-and-unifying-graph-convolutions | 88.4% | 86.9% | - |
| decoupled-spatial-temporal-attention-network | 89.0 % | 86.6% | - |
| maskclr-attention-guided-contrastive-learning | 89.5 | 87.4 | - |
| hyperbolic-self-paced-learning-for-self | 82 | 81.4 | - |
| online-skeleton-based-action-recognition-with | 82 | 80.4 | 0.44 |
| global-context-aware-attention-lstm-networks | 59.2% | 58.3% | - |
| online-skeleton-based-action-recognition-with | 81.7 | 79.7 | 0.15 |
| tsgcnext-dynamic-static-multi-graph | 91.7 | 90.2 | - |
| hierarchically-decomposed-graph-convolutional | 91.6 | 90.1 | - |
| masked-motion-predictors-are-strong-3d-action | 91.3 | 90.0 | - |
| constructing-stronger-and-faster-baselines | 88.0 | 87.9 | - |
| skeleton-based-online-action-prediction-using | 62.4% | 59.9% | - |
| richly-activated-graph-convolutional-network | 82.7% | 81.1% | - |
| online-skeleton-based-action-recognition-with | 85.4 | 84 | 37.38 |
| skelemotion-a-new-representation-of-skeleton | 63.0% | 62.9% | - |
| tsgcnext-dynamic-static-multi-graph | 90.3 | 89.1 | - |
| learning-skeletal-graph-neural-networks-for | 89.2 | 87.5 | - |
| online-skeleton-based-action-recognition-with | 79.1 | 77.3 | 0.22 |
| online-skeleton-based-action-recognition-with | 85.1 | 83.7 | 33.46 |
| hypergraph-transformer-for-skeleton-based | 91.3 | 89.9 | - |
| mix-dimension-in-poincare-geometry-for-3d | 83.2% | 80.5% | - |
| psumnet-unified-modality-part-streams-are-all | 90.6 | 89.4 | - |
| language-supervised-training-for-skeleton | 91.1 | 89.9 | - |
| llms-are-good-action-recognizers | 91.5 | 88.7 | - |
| モデル 74 | 86.90 | 84.88 | - |
| blockgcn-redefine-topology-awareness-for | 91.5 | 90.3 | - |
| spatial-temporal-graph-convolutional-networks-1 | 89.0 | 84.7 | - |
| online-skeleton-based-action-recognition-with | 86.2 | 84.8 | 32.4 |
| multi-modality-co-learning-for-efficient-1 | 91.7 | 90.3 | - |
| vpn-learning-video-pose-embedding-for | 87.8 | 86.3 | - |
| stronger-faster-and-more-explainable-a-graph | 88.3 | 87.3 | - |
| quo-vadis-skeleton-action-recognition | 88.8% | 87.22% | - |