Region Ensemble Network: Improving Convolutional Network for Hand Pose Estimation
This addresses hand pose estimation for human-computer interaction, representing an incremental improvement over existing deep convolutional methods.
The paper tackles hand pose estimation from monocular depth images by proposing a Region Ensemble Network (REN) that partitions convolution outputs into regions and integrates multiple regressors, achieving state-of-the-art performance on two public datasets.
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.