An intuitive multi-frequency feature representation for SO(3)-equivariant networks
This work addresses the problem of insufficient feature representation in SO(3)-equivariant networks for 3D vision tasks, offering an incremental improvement for researchers in computer vision.
The paper tackles the limitation of Vector Neuron (VN) networks in 3D vision tasks by introducing a multi-frequency feature representation that maps 3D points to a high-dimensional space, resulting in VN capturing more details and overcoming its original performance constraints.
The usage of 3D vision algorithms, such as shape reconstruction, remains limited because they require inputs to be at a fixed canonical rotation. Recently, a simple equivariant network, Vector Neuron (VN) has been proposed that can be easily used with the state-of-the-art 3D neural network (NN) architectures. However, its performance is limited because it is designed to use only three-dimensional features, which is insufficient to capture the details present in 3D data. In this paper, we introduce an equivariant feature representation for mapping a 3D point to a high-dimensional feature space. Our feature can discern multiple frequencies present in 3D data, which is the key to designing an expressive feature for 3D vision tasks. Our representation can be used as an input to VNs, and the results demonstrate that with our feature representation, VN captures more details, overcoming the limitation raised in its original paper.