MonoScene: Monocular 3D Semantic Scene Completion

TL;DR

MonoScene introduces a novel monocular 3D semantic scene completion framework that infers dense 3D geometry and semantics from a single RGB image, outperforming existing methods and hallucinating beyond the camera view.

Contribution

The paper presents a new monocular SSC framework with a 2D-3D feature projection, a 3D context relation prior, and novel loss functions, advancing the state-of-the-art in 3D scene understanding from monocular images.

Findings

Outperforms existing methods on all metrics and datasets.

Successfully hallucines plausible scene parts beyond the camera view.

Introduces a novel 2D-3D feature projection and spatio-semantic consistency enforcement.

Abstract

MonoScene proposes a 3D Semantic Scene Completion (SSC) framework, where the dense geometry and semantics of a scene are inferred from a single monocular RGB image. Different from the SSC literature, relying on 2.5 or 3D input, we solve the complex problem of 2D to 3D scene reconstruction while jointly inferring its semantics. Our framework relies on successive 2D and 3D UNets bridged by a novel 2D-3D features projection inspiring from optics and introduces a 3D context relation prior to enforce spatio-semantic consistency. Along with architectural contributions, we introduce novel global scene and local frustums losses. Experiments show we outperform the literature on all metrics and datasets while hallucinating plausible scenery even beyond the camera field of view. Our code and trained models are available at https://github.com/cv-rits/MonoScene.