DriveTok: 3D Driving Scene Tokenization for Unified Multi-View Reconstruction and Understanding
This addresses a bottleneck in autonomous driving systems by improving multi-view scene representation, though it is incremental as it builds on existing vision foundation models and tokenization methods.
The paper tackles the inefficiency and inconsistency of existing tokenizers for multi-view driving scenes by proposing DriveTok, a 3D tokenizer that integrates semantic, geometric, and textural information, achieving strong performance on tasks like image reconstruction and semantic segmentation on the nuScenes dataset.
With the growing adoption of vision-language-action models and world models in autonomous driving systems, scalable image tokenization becomes crucial as the interface for the visual modality. However, most existing tokenizers are designed for monocular and 2D scenes, leading to inefficiency and inter-view inconsistency when applied to high-resolution multi-view driving scenes. To address this, we propose DriveTok, an efficient 3D driving scene tokenizer for unified multi-view reconstruction and understanding. DriveTok first obtains semantically rich visual features from vision foundation models and then transforms them into the scene tokens with 3D deformable cross-attention. For decoding, we employ a multi-view transformer to reconstruct multi-view features from the scene tokens and use multiple heads to obtain RGB, depth, and semantic reconstructions. We also add a 3D head directly on the scene tokens for 3D semantic occupancy prediction for better spatial awareness. With the multiple training objectives, DriveTok learns unified scene tokens that integrate semantic, geometric, and textural information for efficient multi-view tokenization. Extensive experiments on the widely used nuScenes dataset demonstrate that the scene tokens from DriveTok perform well on image reconstruction, semantic segmentation, depth prediction, and 3D occupancy prediction tasks.