Skeleton Based Action Recognition On Ntu Rgbd 1

Metrics

Accuracy (Cross-Setup)

Accuracy (Cross-Subject)

GFLOPS per prediction

Results

Performance results of various models on this benchmark

Model Name	Accuracy (Cross-Setup)	Accuracy (Cross-Subject)	GFLOPS per prediction	Paper Title	Repository
S-TR (1-stream)	81.8	80.2	16.2	Continual Spatio-Temporal Graph Convolutional Networks
Dynamic Skeletons	54.7%	50.8%	-	Jointly learning heterogeneous features for rgb-d activity recognition	-
DG-STGCN	91.3	89.6	-	DG-STGCN: Dynamic Spatial-Temporal Modeling for Skeleton-based Action Recognition
JMDA (based on Skeleton MixFormer)	91.9	90.9	-	Joint Mixing Data Augmentation for Skeleton-based Action Recognition
ST-GCN (1-stream)	-	79	16.73	Continual Spatio-Temporal Graph Convolutional Networks
3s-HYSP	86.3	84.5	-	HYperbolic Self-Paced Learning for Self-Supervised Skeleton-based Action Representations
AGCN (1-stream)	80.7	79.7	18.69	Continual Spatio-Temporal Graph Convolutional Networks
SkeletonGCL (based on CTR-GCN)	91.0	89.5	-	Graph Contrastive Learning for Skeleton-based Action Recognition
EfficientGCN-B0	84.3	85.9	-	Constructing Stronger and Faster Baselines for Skeleton-based Action Recognition
TSRJI (Late Fusion) + HCN	62.8%	67.9%	-	Skeleton Image Representation for 3D Action Recognition based on Tree Structure and Reference Joints
Two-Stream Attention LSTM	63.3%	61.2%	-	Skeleton-Based Human Action Recognition with Global Context-Aware Attention LSTM Networks	-
EfficientGCN-B4	89.1	88.7	-	Constructing Stronger and Faster Baselines for Skeleton-based Action Recognition
Spatio-Temporal LSTM	57.9%	55.7%	-	Spatio-Temporal LSTM with Trust Gates for 3D Human Action Recognition	-
GVFE + AS-GCN with DH-TCN	79.8%	78.3%	-	Vertex Feature Encoding and Hierarchical Temporal Modeling in a Spatial-Temporal Graph Convolutional Network for Action Recognition	-
Multi-Task Learning Network	57.9%	58.4%	-	A New Representation of Skeleton Sequences for 3D Action Recognition	-
TCA-GCN	90.8	89.4	-	Skeleton-based Action Recognition via Temporal-Channel Aggregation
DualHead-Net	89.3	88.2	-	Learning Multi-Granular Spatio-Temporal Graph Network for Skeleton-based Action Recognition
SkateFormer	91.4	89.8	-	SkateFormer: Skeletal-Temporal Transformer for Human Action Recognition
InfoGCN	91.2	89.8	-	InfoGCN: Representation Learning for Human Skeleton-Based Action Recognition
Gimme Signals (Skeleton, AIS)	71.6%	70.8%	-	Gimme Signals: Discriminative signal encoding for multimodal activity recognition

0 of 81 row(s) selected.