High-dimensional Convolutional Networks for Geometric Pattern Recognition
This work addresses geometric registration problems in science and engineering, representing an incremental improvement over existing deep learning approaches.
The paper tackles pattern recognition in high-dimensional geometric spaces by introducing high-dimensional convolutional networks (ConvNets) for detecting linear subspaces up to 32 dimensions and applying them to 3D registration and image correspondence, outperforming prior deep network methods based on global pooling operators.
Many problems in science and engineering can be formulated in terms of geometric patterns in high-dimensional spaces. We present high-dimensional convolutional networks (ConvNets) for pattern recognition problems that arise in the context of geometric registration. We first study the effectiveness of convolutional networks in detecting linear subspaces in high-dimensional spaces with up to 32 dimensions: much higher dimensionality than prior applications of ConvNets. We then apply high-dimensional ConvNets to 3D registration under rigid motions and image correspondence estimation. Experiments indicate that our high-dimensional ConvNets outperform prior approaches that relied on deep networks based on global pooling operators.