Label-efficient Semantic Scene Completion with Scribble Annotations

TL;DR

This paper introduces ScribbleSC, a label-efficient benchmark for semantic scene completion using sparse scribble annotations combined with dense geometric labels, and proposes Scribble2Scene, a method that achieves near fully-supervised performance with significantly fewer labels.

Contribution

The paper presents a novel benchmark and an effective method for semantic scene completion that reduces annotation costs by leveraging sparse scribble labels and a distillation approach.

Findings

Scribble2Scene achieves 99% of fully-supervised performance.

Only 13.5% of voxels need to be labeled for competitive results.

The approach significantly reduces labeling effort in semantic scene completion.

Abstract

Semantic scene completion aims to infer the 3D geometric structures with semantic classes from camera or LiDAR, which provide essential occupancy information in autonomous driving. Prior endeavors concentrate on constructing the network or benchmark in a fully supervised manner. While the dense occupancy grids need point-wise semantic annotations, which incur expensive and tedious labeling costs. In this paper, we build a new label-efficient benchmark, named ScribbleSC, where the sparse scribble-based semantic labels are combined with dense geometric labels for semantic scene completion. In particular, we propose a simple yet effective approach called Scribble2Scene, which bridges the gap between the sparse scribble annotations and fully-supervision. Our method consists of geometric-aware auto-labelers construction and online model training with an offline-to-online distillation module to enhance the performance. Experiments on SemanticKITTI demonstrate that Scribble2Scene achieves competitive performance against the fully-supervised counterparts, showing 99% performance of the fully-supervised models with only 13.5% voxels labeled. Both annotations of ScribbleSC and our full implementation are available at https://github.com/songw-zju/Scribble2Scene.