Region Ensemble Network: Improving Convolutional Network for Hand Pose Estimation

Abstract

Hand pose estimation from monocular depth images is an important and challenging problem for human-computer interaction. Recently deep convolutional networks (ConvNet) with sophisticated design have been employed to address it, but the improvement over traditional methods is not so apparent. To promote the performance of directly 3D coordinate regression, we propose a tree-structured Region Ensemble Network (REN), which partitions the convolution outputs into regions and integrates the results from multiple regressors on each regions. Compared with multi-model ensemble, our model is completely end-to-end training. The experimental results demonstrate that our approach achieves the best performance among state-of-the-arts on two public datasets.

arXiv commentAccepted to ICIP 2017. Project: https://github.com/guohengkai/region-ensemble-network

References in corpus (2)

• Human pose estimation via Convolutional Part Heatmap Regression
• DeepTrack: Learning Discriminative Feature Representations Online for Robust Visual Tracking

Cited by in corpus (6)

• PEPSI++: Fast and Lightweight Network for Image Inpainting
• Joint Hand-object 3D Reconstruction from a Single Image with Cross-branch Feature Fusion
• Multi-Domain Pose Network for Multi-Person Pose Estimation and Tracking
• Pixel-wise Regression: 3D Hand Pose Estimation via Spatial-form Representation and Differentiable Decoder
• Context-Aware Deep Spatio-Temporal Network for Hand Pose Estimation from Depth Images
• Fast and Accurate 3D Hand Pose Estimation via Recurrent Neural Network for Capturing Hand Articulations