CIRCLE: Convolutional Implicit Reconstruction and Completion for Large-scale Indoor Scene

TL;DR

CIRCLE is a fast, efficient framework that uses local implicit functions and a differentiable rendering module to accurately complete and refine large-scale indoor 3D scenes, preserving details and structural integrity.

Contribution

It introduces a novel end-to-end sparse convolutional network with a differentiable rendering module for improved scene completion and refinement.

Findings

Achieves better reconstruction quality than existing methods.

Operates 10-50x faster than closest competitors.

Effectively preserves fine-grained details in large-scale scenes.

Abstract

We present CIRCLE, a framework for large-scale scene completion and geometric refinement based on local implicit signed distance functions. It is based on an end-to-end sparse convolutional network, CircNet, that jointly models local geometric details and global scene structural contexts, allowing it to preserve fine-grained object detail while recovering missing regions commonly arising in traditional 3D scene data. A novel differentiable rendering module enables test-time refinement for better reconstruction quality. Extensive experiments on both real-world and synthetic datasets show that our concise framework is efficient and effective, achieving better reconstruction quality than the closest competitor while being 10-50x faster.