A Coarse-to-Fine Approach to Multi-Modality 3D Occupancy Grounding
By: Zhan Shi , Song Wang , Junbo Chen and more
Potential Business Impact:
Enables self-driving cars to spot objects precisely in 3D
Visual grounding aims to identify objects or regions in a scene based on natural language descriptions, essential for spatially aware perception in autonomous driving. However, existing visual grounding tasks typically depend on bounding boxes that often fail to capture fine-grained details. Not all voxels within a bounding box are occupied, resulting in inaccurate object representations. To address this, we introduce a benchmark for 3D occupancy grounding in challenging outdoor scenes. Built on the nuScenes dataset, it integrates natural language with voxel-level occupancy annotations, offering more precise object perception compared to the traditional grounding task. Moreover, we propose GroundingOcc, an end-to-end model designed for 3D occupancy grounding through multi-modal learning. It combines visual, textual, and point cloud features to predict object location and occupancy information from coarse to fine. Specifically, GroundingOcc comprises a multimodal encoder for feature extraction, an occupancy head for voxel-wise predictions, and a grounding head to refine localization. Additionally, a 2D grounding module and a depth estimation module enhance geometric understanding, thereby boosting model performance. Extensive experiments on the benchmark demonstrate that our method outperforms existing baselines on 3D occupancy grounding. The dataset is available at https://github.com/RONINGOD/GroundingOcc.
Similar Papers
A Coarse-to-Fine Approach to Multi-Modality 3D Occupancy Grounding
CV and Pattern Recognition
Lets self-driving cars spot exact 3D object shapes
QueryOcc: Query-based Self-Supervision for 3D Semantic Occupancy
CV and Pattern Recognition
Teaches cars to see and understand 3D worlds.
ChangingGrounding: 3D Visual Grounding in Changing Scenes
CV and Pattern Recognition
Robots find things in changing rooms using memory.