QueryOcc: Query-based Self-Supervision for 3D Semantic Occupancy

TL;DR

QueryOcc introduces a novel query-based self-supervised framework for learning continuous 3D semantic occupancy directly from sensor data, improving spatial precision and scalability for autonomous driving applications.

Contribution

It presents a new 4D query-based approach with a contractive scene representation for efficient, long-range 3D semantic occupancy learning without manual labels.

Findings

Surpasses previous camera-based methods by 26% in semantic RayIoU.

Operates at 11.6 FPS, enabling real-time applications.

Supports supervision from pseudo-point clouds or raw lidar data.

Abstract

Learning 3D scene geometry and semantics from images is a core challenge in computer vision and a key capability for autonomous driving. Since large-scale 3D annotation is prohibitively expensive, recent work explores self-supervised learning directly from sensor data without manual labels. Existing approaches either rely on 2D rendering consistency, where 3D structure emerges only implicitly, or on discretized voxel grids from accumulated lidar point clouds, limiting spatial precision and scalability. We introduce QueryOcc, a query-based self-supervised framework that learns continuous 3D semantic occupancy directly through independent 4D spatio-temporal queries sampled across adjacent frames. The framework supports supervision from either pseudo-point clouds derived from vision foundation models or raw lidar data. To enable long-range supervision and reasoning under constant memory, we introduce a contractive scene representation that preserves near-field detail while smoothly compressing distant regions. QueryOcc surpasses previous camera-based methods by 26% in semantic RayIoU on the self-supervised Occ3D-nuScenes benchmark while running at 11.6 FPS, demonstrating that direct 4D query supervision enables strong self-supervised occupancy learning. https://research.zenseact.com/publications/queryocc/