Skeleton Based Action Recognition
Skeleton-based Action Recognition is a computer vision task that focuses on recognizing and classifying human actions from sequences of 3D skeletal joint data captured by sensors. The goal of this task is to develop algorithms capable of understanding changes in human posture and accurately determining the type of action, with broad application prospects including human-computer interaction, motion analysis, and security monitoring.
NTU RGB+D
PoseC3D [3D Heatmap]
NTU RGB+D 120
CTR-GCN
Kinetics-Skeleton dataset
PoseC3D (SlowOnly-346)
N-UCLA
SGN
J-HMDB
SBU / SBU-Refine
Joint Line Distance
SYSU 3D
SGN
UAV-Human
HDBN
CAD-120
Florence 3D
UT-Kinect
Temporal Subspace Clustering
Varying-view RGB-D Action-Skeleton
JHMDB (2D poses only)
DD-Net
SHREC 2017 track on 3D Hand Gesture Recognition
TD-GCN
First-Person Hand Action Benchmark
TCN-Summ
Gaming 3D (G3D)
H2O (2 Hands and Objects)
ISTA-Net
MSR Action3D
Temporal K-Means Clustering + Temporal Subspace Clustering
PKU-MMD
JHMDB Pose Tracking
mgPFF+ft 1st
NTU60-X
4s-ShiftGCN
UPenn Action
UNIK
UWA3D
VA-fusion (aug.)
Drive&Act
dyalyt
J-HMBD Early Action
DR^2N
MSRC-12
TCG-dataset
Bidirectional LSTM
Skeletics-152
4s-ShiftGCN
HDM05
HMDB51
Kinetics-400
STGAT
MSR ActionPairs
Temporal Subspace Clustering
Skeleton-Mimetics
MS-G3D
UCF101