UniPlane: Unified Plane Detection and Reconstruction from Posed Monocular Videos

TL;DR

UniPlane introduces a unified Transformer-based neural network that simultaneously detects and reconstructs 3D planes from posed monocular videos, leveraging temporal information for improved accuracy.

Contribution

It unifies plane detection and reconstruction into a single network, directly optimizing for reconstruction quality and utilizing temporal data effectively.

Findings

Outperforms state-of-the-art in plane detection and reconstruction

Achieves +4.6 F-score improvement in geometry detection

Demonstrates robust performance on real-world datasets

Abstract

We present UniPlane, a novel method that unifies plane detection and reconstruction from posed monocular videos. Unlike existing methods that detect planes from local observations and associate them across the video for the final reconstruction, UniPlane unifies both the detection and the reconstruction tasks in a single network, which allows us to directly optimize final reconstruction quality and fully leverage temporal information. Specifically, we build a Transformers-based deep neural network that jointly constructs a 3D feature volume for the environment and estimates a set of per-plane embeddings as queries. UniPlane directly reconstructs the 3D planes by taking dot products between voxel embeddings and the plane embeddings followed by binary thresholding. Extensive experiments on real-world datasets demonstrate that UniPlane outperforms state-of-the-art methods in both plane detection and reconstruction tasks, achieving +4.6 in F-score in geometry as well as consistent improvements in other geometry and segmentation metrics.