Learning 3D Equivariant Implicit Function with Patch-Level Pose-Invariant Representation
Keywords: Equivariant implicit neural representation; Pose-invariant representation; Generalizability to 3D objects; Robustness to transformation
TL;DR: We design the patch-level pose-invariant 3D feature representation to represent the 3D shape, resulting in the implicit displacement estimation of 3D query points based on the local patch-level pose-invariant representation.
Abstract: Implicit neural representation gains popularity in modeling the continuous 3D surface for 3D representation and reconstruction. In this work, we are motivated by the fact that the local 3D patches repeatedly appear on 3D shapes/surfaces if the factor of poses is removed. Based on this observation, we propose the 3D patch-level equivariant implicit function (PEIF) based on the 3D patch-level pose-invariant representation, allowing us to reconstruct 3D surfaces by estimating equivariant displacement vector fields for query points. Specifically, our model is based on the pose-normalized query/patch pairs and enhanced by the proposed intrinsic patch geometry representation, modeling the intrinsic 3D patch geometry feature by learnable multi-head memory banks. Extensive experiments show that our model achieves state-of-the-art performance on multiple surface reconstruction datasets, and also exhibits better generalization to crossdataset shapes and robustness to arbitrary rotations. Our code will be available at https://github.com/mathXin112/PEIF.git.
Primary Area: Machine vision
Submission Number: 2801
Loading