TriHorn-Net: A model for accurate depth-based 3D hand pose estimation

Mohammad Rezaei, Razieh Rastgoo, Vassilis Athitsos

Abstract: Highlights•3D hand pose estimation has made significant progress recently.•We propose a model, TriHorn-Net, for accurate depth-based 3D hand pose estimation.•Our model, decomposes the 3D hand pose into the 2D joint locations.•The proposed model uses the Pix-Dropout for appearance-based data augmentation.•The proposed model outperforms the state-of-the-art methods on three public datasets.