Activity Recognition on NTU RGB+D 120

Metric: Accuracy (Cross-Subject) (higher is better)

LeaderboardDataset

Loading chart...

Results

Hide extra data

Sort:

#	Model↕	Accuracy (Cross-Subject)▼	Extra Data	Paper	Date↕	Code
1	DSCNet (RGB + Pose)	95.6	No	-	-	Code
2	PoseC3D (RGB + Pose)	95.3	No	Revisiting Skeleton-based Action Recognition	2021-04-28	Code
3	π-ViT (RGB + Pose)	95.1	Yes	Just Add $π$! Pose Induced Video Transformers fo...	2023-11-30	Code
4	EPAM-Net	94.3	No	EPAM-Net: An Efficient Pose-driven Attention-gui...	2024-08-10	Code
5	MMNet (RGB + Pose)	92.9	Yes	-	-	Code
6	π-ViT (RGB only)	92.9	Yes	Just Add $π$! Pose Induced Video Transformers fo...	2023-11-30	Code
7	VPN++ (RGB + Pose)	92.5	Yes	VPN++: Rethinking Video-Pose embeddings for unde...	2021-05-17	Code
8	DVANet (RGB only)	91.6	No	DVANet: Disentangling View and Action Features f...	2023-12-10	Code
9	EPP-Net (Parsing + Pose)	91.1	No	Explore Human Parsing Modality for Action Recogn...	2024-01-04	Code
10	ProtoGCN	90.9	No	Revealing Key Details to See Differences: A Nove...	2024-11-28	Code
11	JMDA (based on Skeleton MixFormer)	90.9	No	-	-	Code
12	LA-GCN	90.7	Yes	Language Knowledge-Assisted Representation Learn...	2023-05-21	Code
13	MSA-GCN	90.6	No	-	-	-
14	3DA (RGB + Pose)	90.5	No	Cross-Modal Learning with 3D Deformable Attentio...	2022-12-12	-
15	Shap-Mix	90.4	No	Shap-Mix: Shapley Value Guided Mixing for Long-T...	2024-07-17	Code
16	STAR-Transformer (RGB + Pose)	90.3	No	STAR-Transformer: A Spatio-temporal Cross Attent...	2022-10-14	-
17	MMCL	90.3	Yes	Multi-Modality Co-Learning for Efficient Skeleto...	2024-07-22	Code
18	BlockGCN	90.3	No	-	-	Code
19	TSGCNeXt	90.2	No	TSGCNeXt: Dynamic-Static Multi-Graph Convolution...	2023-04-23	Code
20	HD-GCN	90.1	No	Hierarchically Decomposed Graph Convolutional Ne...	2022-08-23	Code
21	IPP-Net (Parsing + Pose)	90	No	Integrating Human Parsing and Pose Network for H...	2023-07-16	Code
22	MAMP	90	No	Masked Motion Predictors are Strong 3D Action Re...	2023-08-14	Code
23	STEP-CATFormer	90	Yes	STEP CATFormer: Spatial-Temporal Effective Body-...	2023-12-06	Code
24	Hyperformer	89.9	No	Hypergraph Transformer for Skeleton-based Action...	2022-11-17	Code
25	LST	89.9	Yes	Generative Action Description Prompts for Skelet...	2022-08-10	Code
26	SkateFormer	89.8	No	SkateFormer: Skeletal-Temporal Transformer for H...	2024-03-14	Code
27	InfoGCN	89.8	No	-	-	Code
28	Stream-GCN	89.7	No	Action Recognition with Multi-stream Motion Mode...	2023-06-13	-
29	DG-STGCN	89.6	No	DG-STGCN: Dynamic Spatial-Temporal Modeling for ...	2022-10-12	Code
30	SkeletonGCL (based on CTR-GCN)	89.5	No	Graph Contrastive Learning for Skeleton-based Ac...	2023-01-26	Code
31	JPFormer(Pose)	89.4	No	-	-	Code
32	JPFormer	89.4	No	-	-	Code
33	TCA-GCN	89.4	No	Skeleton-based Action Recognition via Temporal-C...	2022-05-31	Code
34	PSUMNet	89.4	No	PSUMNet: Unified Modality Part Streams are All Y...	2022-08-11	Code
35	DSTSA-GCN	89.12	No	DSTSA-GCN: Advancing Skeleton-Based Gesture Reco...	2025-01-21	Code
36	DSTSA-GCN	89.12	No	DSTSA-GCN: Advancing Skeleton-Based Gesture Reco...	2025-01-21	Code
37	TSGCNeXT	89.1	No	TSGCNeXt: Dynamic-Static Multi-Graph Convolution...	2023-04-23	Code
38	CTR-GCN	88.9	No	Channel-wise Topology Refinement Graph Convoluti...	2021-07-26	Code
39	Lit-llama	88.7	No	LLMs are Good Action Recognizers	2024-03-31	-
40	STGAT	88.7	No	Spatial Temporal Graph Attention Network for Ske...	2022-08-18	Code
41	EfficientGCN-B4	88.7	No	Constructing Stronger and Faster Baselines for S...	2021-06-29	Code
42	ST-GCN++ [PYSKL, 3D Skeleton]	88.6	No	PYSKL: Towards Good Practices for Skeleton Actio...	2022-05-19	Code
43	DualHead-Net	88.2	No	Learning Multi-Granular Spatio-Temporal Graph Ne...	2021-08-10	Code
44	EfficientGCN-B2	87.9	No	Constructing Stronger and Faster Baselines for S...	2021-06-29	Code
45	VPN (RGB + Pose)	87.8	Yes	VPN: Learning Video-Pose Embedding for Activitie...	2020-07-06	Code
46	Skeletal GNN	87.5	No	Learning Skeletal Graph Neural Networks for Hard...	2021-08-16	-
47	MaskCLR	87.4	No	-	-	-
48	MSSTNet	87.4	No	-	-	Code
49	PA-ResGCN-B19	87.3	No	Stronger, Faster and More Explainable: A Graph C...	2020-10-20	Code
50	PoseC3D (w. HRNet 2D Skeleton)	86.9	No	Revisiting Skeleton-based Action Recognition	2021-04-28	Code
51	VPN	86.3	Yes	VPN: Learning Video-Pose Embedding for Activitie...	2020-07-06	Code
52	ST-GCN [PYSKL, 3D Skeleton]	86.2	No	Spatial Temporal Graph Convolutional Networks fo...	2018-01-23	Code
53	EfficientGCN-B0	85.9	No	Constructing Stronger and Faster Baselines for S...	2021-06-29	Code
54	ViewCon (RGB)	85.6	No	-	-	Code
55	VA-CNN (ResNeXt-101)	84.88	No	-	-	-
56	S-TR (2-stream)	84.8	No	Continual Spatio-Temporal Graph Convolutional Ne...	2022-03-21	Code
57	CoS-TR* (2-stream)	84.8	No	Continual Spatio-Temporal Graph Convolutional Ne...	2022-03-21	Code
58	ST-GCN [PYSKL, 2D Skeleton]	84.7	No	Spatial Temporal Graph Convolutional Networks fo...	2018-01-23	Code
59	3s-HYSP	84.5	No	HYperbolic Self-Paced Learning for Self-Supervis...	2023-03-10	Code
60	CoST-GCN* (2-stream)	84	No	Continual Spatio-Temporal Graph Convolutional Ne...	2022-03-21	Code
61	AGCN (2-stream)	84	No	Continual Spatio-Temporal Graph Convolutional Ne...	2022-03-21	Code
62	ST-GCN (2-stream)	83.7	No	Continual Spatio-Temporal Graph Convolutional Ne...	2022-03-21	Code
63	HYSP	81.4	No	HYperbolic Self-Paced Learning for Self-Supervis...	2023-03-10	Code
64	3s-USDRL (DSTE) This work	80.6	No	USDRL: Unified Skeleton-Based Dense Representati...	2024-12-12	Code
65	CoAGCN* (2-stream)	80.4	No	Continual Spatio-Temporal Graph Convolutional Ne...	2022-03-21	Code
66	S-TR (1-stream)	80.2	No	Continual Spatio-Temporal Graph Convolutional Ne...	2022-03-21	Code
67	CoS-TR* (1-stream)	79.7	No	Continual Spatio-Temporal Graph Convolutional Ne...	2022-03-21	Code
68	AGCN (1-stream)	79.7	No	Continual Spatio-Temporal Graph Convolutional Ne...	2022-03-21	Code
69	CoST-GCN* (1-stream)	79.4	No	Continual Spatio-Temporal Graph Convolutional Ne...	2022-03-21	Code
70	ST-GCN + AS-GCN w/DH-TCN	79.2	No	Vertex Feature Encoding and Hierarchical Tempora...	2019-12-20	-
71	ST-GCN (1-stream)	79	No	Continual Spatio-Temporal Graph Convolutional Ne...	2022-03-21	Code
72	CoAGCN* (1-stream)	77.3	No	Continual Spatio-Temporal Graph Convolutional Ne...	2022-03-21	Code
73	Gimme Signals (AIS)	71.59	No	Gimme Signals: Discriminative signal encoding fo...	2020-03-13	Code
74	Skelemotion + Yang et al. (skeleton only)	67.7	No	SkeleMotion: A New Representation of Skeleton Jo...	2019-07-30	Code
75	Body Pose Evolution Map	66.9	No	-	-	-
76	TSRJI	62.8	No	Skeleton Image Representation for 3D Action Reco...	2019-09-11	Code

#1DSCNet (RGB + Pose)
95.6
Accuracy (Cross-Subject)
No paperCode
#2PoseC3D (RGB + Pose)SOTA
95.3
Accuracy (Cross-Subject)· 2021-04-28
Revisiting Skeleton-based Action Recognition Code
#3π-ViT (RGB + Pose)
95.1
Accuracy (Cross-Subject)· Extra Data· 2023-11-30
Just Add $π$! Pose Induced Video Transformers for Understanding Activities of Daily Living Code
#4EPAM-Net
94.3
Accuracy (Cross-Subject)· 2024-08-10
EPAM-Net: An Efficient Pose-driven Attention-guided Multimodal Network for Video Action Recognition Code
#5MMNet (RGB + Pose)
92.9
Accuracy (Cross-Subject)· Extra Data
No paperCode
#6π-ViT (RGB only)
92.9
Accuracy (Cross-Subject)· Extra Data· 2023-11-30
Just Add $π$! Pose Induced Video Transformers for Understanding Activities of Daily Living Code
#7VPN++ (RGB + Pose)
92.5
Accuracy (Cross-Subject)· Extra Data· 2021-05-17
VPN++: Rethinking Video-Pose embeddings for understanding Activities of Daily Living Code
#8DVANet (RGB only)
91.6
Accuracy (Cross-Subject)· 2023-12-10
DVANet: Disentangling View and Action Features for Multi-View Action Recognition Code
#9EPP-Net (Parsing + Pose)
91.1
Accuracy (Cross-Subject)· 2024-01-04
Explore Human Parsing Modality for Action Recognition Code
#10ProtoGCN
90.9
Accuracy (Cross-Subject)· 2024-11-28
Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition Code
#11JMDA (based on Skeleton MixFormer)
90.9
Accuracy (Cross-Subject)
No paperCode
#12LA-GCN
90.7
Accuracy (Cross-Subject)· Extra Data· 2023-05-21
Language Knowledge-Assisted Representation Learning for Skeleton-Based Action Recognition Code
#13MSA-GCN
90.6
Accuracy (Cross-Subject)
No paper
#143DA (RGB + Pose)
90.5
Accuracy (Cross-Subject)· 2022-12-12
Cross-Modal Learning with 3D Deformable Attention for Action Recognition
#15Shap-Mix
90.4
Accuracy (Cross-Subject)· 2024-07-17
Shap-Mix: Shapley Value Guided Mixing for Long-Tailed Skeleton Based Action Recognition Code
#16STAR-Transformer (RGB + Pose)
90.3
Accuracy (Cross-Subject)· 2022-10-14
STAR-Transformer: A Spatio-temporal Cross Attention Transformer for Human Action Recognition
#17MMCL
90.3
Accuracy (Cross-Subject)· Extra Data· 2024-07-22
Multi-Modality Co-Learning for Efficient Skeleton-based Action Recognition Code
#18BlockGCN
90.3
Accuracy (Cross-Subject)
No paperCode
#19TSGCNeXt
90.2
Accuracy (Cross-Subject)· 2023-04-23
TSGCNeXt: Dynamic-Static Multi-Graph Convolution for Efficient Skeleton-Based Action Recognition with Long-term Learning Potential Code
#20HD-GCN
90.1
Accuracy (Cross-Subject)· 2022-08-23
Hierarchically Decomposed Graph Convolutional Networks for Skeleton-Based Action Recognition Code
#21IPP-Net (Parsing + Pose)
90
Accuracy (Cross-Subject)· 2023-07-16
Integrating Human Parsing and Pose Network for Human Action Recognition Code
#22MAMP
90
Accuracy (Cross-Subject)· 2023-08-14
Masked Motion Predictors are Strong 3D Action Representation Learners Code
#23STEP-CATFormer
90
Accuracy (Cross-Subject)· Extra Data· 2023-12-06
STEP CATFormer: Spatial-Temporal Effective Body-Part Cross Attention Transformer for Skeleton-based Action Recognition Code
#24Hyperformer
89.9
Accuracy (Cross-Subject)· 2022-11-17
Hypergraph Transformer for Skeleton-based Action Recognition Code
#25LST
89.9
Accuracy (Cross-Subject)· Extra Data· 2022-08-10
Generative Action Description Prompts for Skeleton-based Action Recognition Code
#26SkateFormer
89.8
Accuracy (Cross-Subject)· 2024-03-14
SkateFormer: Skeletal-Temporal Transformer for Human Action Recognition Code
#27InfoGCN
89.8
Accuracy (Cross-Subject)
No paperCode
#28Stream-GCN
89.7
Accuracy (Cross-Subject)· 2023-06-13
Action Recognition with Multi-stream Motion Modeling and Mutual Information Maximization
#29DG-STGCN
89.6
Accuracy (Cross-Subject)· 2022-10-12
DG-STGCN: Dynamic Spatial-Temporal Modeling for Skeleton-based Action Recognition Code
#30SkeletonGCL (based on CTR-GCN)
89.5
Accuracy (Cross-Subject)· 2023-01-26
Graph Contrastive Learning for Skeleton-based Action Recognition Code
#31JPFormer(Pose)
89.4
Accuracy (Cross-Subject)
No paperCode
#32JPFormer
89.4
Accuracy (Cross-Subject)
No paperCode
#33TCA-GCN
89.4
Accuracy (Cross-Subject)· 2022-05-31
Skeleton-based Action Recognition via Temporal-Channel Aggregation Code
#34PSUMNet
89.4
Accuracy (Cross-Subject)· 2022-08-11
PSUMNet: Unified Modality Part Streams are All You Need for Efficient Pose-based Action Recognition Code
#35DSTSA-GCN
89.12
Accuracy (Cross-Subject)· 2025-01-21
DSTSA-GCN: Advancing Skeleton-Based Gesture Recognition with Semantic-Aware Spatio-Temporal Topology Modeling Code
#36DSTSA-GCN
89.12
Accuracy (Cross-Subject)· 2025-01-21
DSTSA-GCN: Advancing Skeleton-Based Gesture Recognition with Semantic-Aware Spatio-Temporal Topology Modeling Code
#37TSGCNeXT
89.1
Accuracy (Cross-Subject)· 2023-04-23
TSGCNeXt: Dynamic-Static Multi-Graph Convolution for Efficient Skeleton-Based Action Recognition with Long-term Learning Potential Code
#38CTR-GCN
88.9
Accuracy (Cross-Subject)· 2021-07-26
Channel-wise Topology Refinement Graph Convolution for Skeleton-Based Action Recognition Code
#39Lit-llama
88.7
Accuracy (Cross-Subject)· 2024-03-31
LLMs are Good Action Recognizers
#40STGAT
88.7
Accuracy (Cross-Subject)· 2022-08-18
Spatial Temporal Graph Attention Network for Skeleton-Based Action Recognition Code
#41EfficientGCN-B4
88.7
Accuracy (Cross-Subject)· 2021-06-29
Constructing Stronger and Faster Baselines for Skeleton-based Action Recognition Code
#42ST-GCN++ [PYSKL, 3D Skeleton]
88.6
Accuracy (Cross-Subject)· 2022-05-19
PYSKL: Towards Good Practices for Skeleton Action Recognition Code
#43DualHead-Net
88.2
Accuracy (Cross-Subject)· 2021-08-10
Learning Multi-Granular Spatio-Temporal Graph Network for Skeleton-based Action Recognition Code
#44EfficientGCN-B2
87.9
Accuracy (Cross-Subject)· 2021-06-29
Constructing Stronger and Faster Baselines for Skeleton-based Action Recognition Code
#45VPN (RGB + Pose)SOTA
87.8
Accuracy (Cross-Subject)· Extra Data· 2020-07-06
VPN: Learning Video-Pose Embedding for Activities of Daily Living Code
#46Skeletal GNN
87.5
Accuracy (Cross-Subject)· 2021-08-16
Learning Skeletal Graph Neural Networks for Hard 3D Pose Estimation
#47MaskCLR
87.4
Accuracy (Cross-Subject)
No paper
#48MSSTNet
87.4
Accuracy (Cross-Subject)
No paperCode
#49PA-ResGCN-B19
87.3
Accuracy (Cross-Subject)· 2020-10-20
Stronger, Faster and More Explainable: A Graph Convolutional Baseline for Skeleton-based Action Recognition Code
#50PoseC3D (w. HRNet 2D Skeleton)
86.9
Accuracy (Cross-Subject)· 2021-04-28
Revisiting Skeleton-based Action Recognition Code
#51VPN
86.3
Accuracy (Cross-Subject)· Extra Data· 2020-07-06
VPN: Learning Video-Pose Embedding for Activities of Daily Living Code
#52ST-GCN [PYSKL, 3D Skeleton]SOTA
86.2
Accuracy (Cross-Subject)· 2018-01-23
Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition Code
#53EfficientGCN-B0
85.9
Accuracy (Cross-Subject)· 2021-06-29
Constructing Stronger and Faster Baselines for Skeleton-based Action Recognition Code
#54ViewCon (RGB)
85.6
Accuracy (Cross-Subject)
No paperCode
#55VA-CNN (ResNeXt-101)
84.88
Accuracy (Cross-Subject)
No paper
#56S-TR (2-stream)
84.8
Accuracy (Cross-Subject)· 2022-03-21
Continual Spatio-Temporal Graph Convolutional Networks Code
#57CoS-TR* (2-stream)
84.8
Accuracy (Cross-Subject)· 2022-03-21
Continual Spatio-Temporal Graph Convolutional Networks Code
#58ST-GCN [PYSKL, 2D Skeleton]
84.7
Accuracy (Cross-Subject)· 2018-01-23
Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition Code
#593s-HYSP
84.5
Accuracy (Cross-Subject)· 2023-03-10
HYperbolic Self-Paced Learning for Self-Supervised Skeleton-based Action Representations Code
#60CoST-GCN* (2-stream)
84
Accuracy (Cross-Subject)· 2022-03-21
Continual Spatio-Temporal Graph Convolutional Networks Code
#61AGCN (2-stream)
84
Accuracy (Cross-Subject)· 2022-03-21
Continual Spatio-Temporal Graph Convolutional Networks Code
#62ST-GCN (2-stream)
83.7
Accuracy (Cross-Subject)· 2022-03-21
Continual Spatio-Temporal Graph Convolutional Networks Code
#63HYSP
81.4
Accuracy (Cross-Subject)· 2023-03-10
HYperbolic Self-Paced Learning for Self-Supervised Skeleton-based Action Representations Code
#643s-USDRL (DSTE) This work
80.6
Accuracy (Cross-Subject)· 2024-12-12
USDRL: Unified Skeleton-Based Dense Representation Learning with Multi-Grained Feature Decorrelation Code
#65CoAGCN* (2-stream)
80.4
Accuracy (Cross-Subject)· 2022-03-21
Continual Spatio-Temporal Graph Convolutional Networks Code
#66S-TR (1-stream)
80.2
Accuracy (Cross-Subject)· 2022-03-21
Continual Spatio-Temporal Graph Convolutional Networks Code
#67CoS-TR* (1-stream)
79.7
Accuracy (Cross-Subject)· 2022-03-21
Continual Spatio-Temporal Graph Convolutional Networks Code
#68AGCN (1-stream)
79.7
Accuracy (Cross-Subject)· 2022-03-21
Continual Spatio-Temporal Graph Convolutional Networks Code
#69CoST-GCN* (1-stream)
79.4
Accuracy (Cross-Subject)· 2022-03-21
Continual Spatio-Temporal Graph Convolutional Networks Code
#70ST-GCN + AS-GCN w/DH-TCN
79.2
Accuracy (Cross-Subject)· 2019-12-20
Vertex Feature Encoding and Hierarchical Temporal Modeling in a Spatial-Temporal Graph Convolutional Network for Action Recognition
#71ST-GCN (1-stream)
79
Accuracy (Cross-Subject)· 2022-03-21
Continual Spatio-Temporal Graph Convolutional Networks Code
#72CoAGCN* (1-stream)
77.3
Accuracy (Cross-Subject)· 2022-03-21
Continual Spatio-Temporal Graph Convolutional Networks Code
#73Gimme Signals (AIS)
71.59
Accuracy (Cross-Subject)· 2020-03-13
Gimme Signals: Discriminative signal encoding for multimodal activity recognition Code
#74Skelemotion + Yang et al. (skeleton only)
67.7
Accuracy (Cross-Subject)· 2019-07-30
SkeleMotion: A New Representation of Skeleton Joint Sequences Based on Motion Information for 3D Action Recognition Code
#75Body Pose Evolution Map
66.9
Accuracy (Cross-Subject)
No paper
#76TSRJI
62.8
Accuracy (Cross-Subject)· 2019-09-11
Skeleton Image Representation for 3D Action Recognition based on Tree Structure and Reference Joints Code