Tasks SotA Datasets Papers Methods Submit About

Papers With Code 2

A community resource for machine learning research: papers, code, benchmarks, and state-of-the-art results.

Explore

Notable Benchmarks All SotA Datasets Papers Methods

Community

Submit Results About

Data sourced from the PWC Archive (CC-BY-SA 4.0). Built by the community, for the community.

Models/MS-G3D

MS-G3D

Reported on 55 benchmarks across 8 tasks · 2 papers · 23 SOTA

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Computer Vision21 results

VideoonNTU60-X
Accuracy (Body + Fingers + Face joints)· 2021-01-27
91.12
SOTA
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
VideoonNTU60-X
Accuracy (Body joints)· 2021-01-27
91.26
SOTA
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
Temporal Action LocalizationonNTU60-X
Accuracy (Body + Fingers + Face joints)· 2021-01-27
91.12
SOTA
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
Temporal Action LocalizationonNTU60-X
Accuracy (Body joints)· 2021-01-27
91.26
SOTA
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
Action LocalizationonNTU60-X
Accuracy (Body + Fingers + Face joints)· 2021-01-27
91.12
SOTA
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
Action LocalizationonNTU60-X
Accuracy (Body joints)· 2021-01-27
91.26
SOTA
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
VideoonAssembly101
Verbs Top-1· 2020-03-31
65.7
best: 69.23 (HandFormer-B/21)
SOTA
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
Temporal Action LocalizationonAssembly101
Verbs Top-1· 2020-03-31
65.7
best: 69.23 (HandFormer-B/21)
SOTA
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
Action LocalizationonAssembly101
Verbs Top-1· 2020-03-31
65.7
best: 69.23 (HandFormer-B/21)
SOTA
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
VideoonNTU60-X
Accuracy (Body + Fingers joints)· 2021-01-27
91.76
best: 91.78 (4s-ShiftGCN)
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
Temporal Action LocalizationonNTU60-X
Accuracy (Body + Fingers joints)· 2021-01-27
91.76
best: 91.78 (4s-ShiftGCN)
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
Action LocalizationonNTU60-X
Accuracy (Body + Fingers joints)· 2021-01-27
91.76
best: 91.78 (4s-ShiftGCN)
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
VideoonAssembly101
Actions Top-1· 2020-03-31
28.7
best: 41.06 (HandFormer-B/21)
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
VideoonAssembly101
Object Top-1· 2020-03-31
36.3
best: 51.17 (HandFormer-B/21)
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
VideoonKinetics-Skeleton dataset
Accuracy· 2020-03-31
38
best: 52.3 (Structured Keypoint Pooling (PPNv2 skeletons+objects))
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
Temporal Action LocalizationonAssembly101
Actions Top-1· 2020-03-31
28.7
best: 41.06 (HandFormer-B/21)
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
Temporal Action LocalizationonAssembly101
Object Top-1· 2020-03-31
36.3
best: 51.17 (HandFormer-B/21)
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
Temporal Action LocalizationonKinetics-Skeleton dataset
Accuracy· 2020-03-31
38
best: 52.3 (Structured Keypoint Pooling (PPNv2 skeletons+objects))
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
Action LocalizationonAssembly101
Actions Top-1· 2020-03-31
28.7
best: 41.06 (HandFormer-B/21)
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
Action LocalizationonAssembly101
Object Top-1· 2020-03-31
36.3
best: 51.17 (HandFormer-B/21)
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
Action LocalizationonKinetics-Skeleton dataset
Accuracy· 2020-03-31
38
best: 52.3 (Structured Keypoint Pooling (PPNv2 skeletons+objects))
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111

Time Series12 results

Action DetectiononNTU60-X
Accuracy (Body + Fingers + Face joints)· 2021-01-27
91.12
SOTA
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
Action DetectiononNTU60-X
Accuracy (Body joints)· 2021-01-27
91.26
SOTA
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
Action RecognitiononNTU60-X
Accuracy (Body + Fingers + Face joints)· 2021-01-27
91.12
SOTA
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
Action RecognitiononNTU60-X
Accuracy (Body joints)· 2021-01-27
91.26
SOTA
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
Action RecognitiononAssembly101
Verbs Top-1· 2020-03-31
65.7
best: 69.23 (HandFormer-B/21)
SOTA
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
Action DetectiononNTU60-X
Accuracy (Body + Fingers joints)· 2021-01-27
91.76
best: 91.78 (4s-ShiftGCN)
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
Action RecognitiononNTU60-X
Accuracy (Body + Fingers joints)· 2021-01-27
91.76
best: 91.78 (4s-ShiftGCN)
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
Action DetectiononKinetics-Skeleton dataset
Accuracy· 2020-03-31
38
best: 52.3 (Structured Keypoint Pooling (PPNv2 skeletons+objects))
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
Action RecognitiononH2O (2 Hands and Objects)
Actions Top-1· 2020-03-31
50.83
best: 93.39 (HandFormer-B/21x8)
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
Action RecognitiononAssembly101
Actions Top-1· 2020-03-31
28.7
best: 41.06 (HandFormer-B/21)
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
Action RecognitiononAssembly101
Object Top-1· 2020-03-31
36.3
best: 51.17 (HandFormer-B/21)
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
Action RecognitiononKinetics-Skeleton dataset
Accuracy· 2020-03-31
38
best: 52.3 (Structured Keypoint Pooling (PPNv2 skeletons+objects))
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111

Robots8 results

Activity RecognitiononNTU60-X
Accuracy (Body + Fingers + Face joints)· 2021-01-27
91.12
SOTA
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
Activity RecognitiononNTU60-X
Accuracy (Body joints)· 2021-01-27
91.26
SOTA
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
Activity RecognitiononAssembly101
Verbs Top-1· 2020-03-31
65.7
best: 69.23 (HandFormer-B/21)
SOTA
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
Activity RecognitiononNTU60-X
Accuracy (Body + Fingers joints)· 2021-01-27
91.76
best: 91.78 (4s-ShiftGCN)
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
Activity RecognitiononH2O (2 Hands and Objects)
Actions Top-1· 2020-03-31
50.83
best: 93.39 (HandFormer-B/21x8)
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
Activity RecognitiononAssembly101
Actions Top-1· 2020-03-31
28.7
best: 41.06 (HandFormer-B/21)
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
Activity RecognitiononAssembly101
Object Top-1· 2020-03-31
36.3
best: 51.17 (HandFormer-B/21)
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
Activity RecognitiononKinetics-Skeleton dataset
Accuracy· 2020-03-31
38
best: 52.3 (Structured Keypoint Pooling (PPNv2 skeletons+objects))
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111

Methodology7 results

Zero-Shot LearningonNTU60-X
Accuracy (Body + Fingers + Face joints)· 2021-01-27
91.12
SOTA
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
Zero-Shot LearningonNTU60-X
Accuracy (Body joints)· 2021-01-27
91.26
SOTA
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
Zero-Shot LearningonAssembly101
Verbs Top-1· 2020-03-31
65.7
best: 69.23 (HandFormer-B/21)
SOTA
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
Zero-Shot LearningonNTU60-X
Accuracy (Body + Fingers joints)· 2021-01-27
91.76
best: 91.78 (4s-ShiftGCN)
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
Zero-Shot LearningonAssembly101
Actions Top-1· 2020-03-31
28.7
best: 41.06 (HandFormer-B/21)
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
Zero-Shot LearningonAssembly101
Object Top-1· 2020-03-31
36.3
best: 51.17 (HandFormer-B/21)
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
Zero-Shot LearningonKinetics-Skeleton dataset
Accuracy· 2020-03-31
38
best: 52.3 (Structured Keypoint Pooling (PPNv2 skeletons+objects))
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111

Natural Language Processing7 results

3D Action RecognitiononNTU60-X
Accuracy (Body + Fingers + Face joints)· 2021-01-27
91.12
SOTA
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
3D Action RecognitiononNTU60-X
Accuracy (Body joints)· 2021-01-27
91.26
SOTA
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
3D Action RecognitiononAssembly101
Verbs Top-1· 2020-03-31
65.7
best: 69.23 (HandFormer-B/21)
SOTA
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
3D Action RecognitiononNTU60-X
Accuracy (Body + Fingers joints)· 2021-01-27
91.76
best: 91.78 (4s-ShiftGCN)
NTU-X: An Enhanced Large-scale Dataset for Improving Pose-based Recognition of Subtle Human Actions arXiv:2101.11529
3D Action RecognitiononAssembly101
Actions Top-1· 2020-03-31
28.7
best: 41.06 (HandFormer-B/21)
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
3D Action RecognitiononAssembly101
Object Top-1· 2020-03-31
36.3
best: 51.17 (HandFormer-B/21)
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111
3D Action RecognitiononKinetics-Skeleton dataset
Accuracy· 2020-03-31
38
best: 52.3 (Structured Keypoint Pooling (PPNv2 skeletons+objects))
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition arXiv:2003.14111