6DCNN with Roto-Translational Convolution Filters for Volumetric Data Processing

In this work, we introduce 6D Convolutional Neural Network (6DCNN) designed to tackle the problem of detecting relative positions and orientations local patterns when processing three-dimensional volumetric data. 6DCNN also includes SE(3)-equivariant message-passing nonlinear activation operations constructed in Fourier space. Working space allows significantly reducing computational complexity our operations. We demonstrate properties convolution its efficiency recognition spatial patterns. assess model on several datasets from recent CASP protein structure prediction challenges. Here, improves over baseline architecture outperforms state art.