Skip to main content
. 2022 May 28;22(11):4109. doi: 10.3390/s22114109
HPE human pose estimation
LSTM long short-term memory
GNN graph neural network
GCN graph convolution network
TCN temporal convolutional network
RNN recurrent neural network
MPJPE mean per joint position error
MRPE mean of the root position error
AUC area under the curve
3D-PCK percentage of correct key-points in 3D space
AProot average precision of the root keypoint
GPU graphics processing unit
GR first 3D absolute pose methodology: GAST-Net + RootNet
GA second 3D absolute pose methodology: GAST-NetABS trained on MuCo-Temp
GAR third 3D absolute pose methodology: GAST-NetABS trained on MuCo-Temp + RootNet
Root-GAST the whole pipeline: human detector + 2D pose estimator + 3D absolute pose estimator