Clustering-Based Refinement for 3D Human Body Parts Segmentation

AbstractA common approach to address human body parts segmentation on 3D data involves the use of a 2D network and projection. Following this approach, several errors could be introduced in final output, such as reprojection errors. Such are even more significant when considering very small hands. In paper, we propose new algorithm that aims reduce improve parts. The detects noise points wrong clusters using DBSCAN algorithm, changes labels exploiting shape position clusters. We evaluated proposed 3DPeople synthetic dataset real dataset, highlighting how it can greatly like With our achieved an improvement up 4.68% IoU 2.30% scenario.KeywordsHuman parsing3D semantic segmentation3D clustering