3D Hand Pose Estimation in Point Cloud Using 3D Convolutional Neural Network on Egocentric Datasets

3D hand pose estimation from egocentric vision is an important study in the construction of assistance systems and modeling robot robotics. In this paper, we propose a complete method for estimating posefrom complex scene data obtained sensor. which simple yet highly efficient pre-processing step segmentation. process, used Hand PointNet (HPN), V2V-PoseNet(V2V), Point-to-Point Regression (PtoP) finetuning to estimate collected sensor, such as CVRA, FPHA (First-Person Action) datasets. HPN, V2V, PtoP are thedeep networks/Convolutional Neural Networks (CNNs) that uses point cloud hand. We evaluate results using preprocessing do not use see effectiveness proposed method. The show distance error increased many times compared estimates on datasets obstructed (the surveillance cameras, viewed top view, front sides view) MSRA, NYU, ICVL quantified, analyzed, shown CVAR dataset projected color image dataset.