# Region Ensemble Network: Improving Convolutional Network for Hand Pose Estimation

## TL;DR

This paper introduces a novel tree-structured Region Ensemble Network (REN) that partitions convolution outputs into regions and integrates multiple regressors, significantly improving 3D hand pose estimation from monocular depth images.

## Contribution

The paper proposes an end-to-end trainable REN that enhances convolutional network performance for hand pose estimation by region-based ensemble methods.

## Key findings

- Achieves state-of-the-art performance on public datasets.
- Outperforms traditional ensemble and single-model approaches.
- Demonstrates effective region-based integration improves accuracy.

## 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.