aiqbench analysis
DVGT introduces a novel transformer architecture for autonomous driving that reconstructs dense 3D point maps from unposed multi-view inputs without relying on explicit camera parameters. By leveraging alternating attention mechanisms across views and frames, the model eliminates the need for post-alignment with external sensors, offering greater flexibility for diverse camera configurations. This approach represents a significant step forward in visual geometry perception, directly addressing the gap for driving-targeted models adaptable to various real-world scenarios.
DVGT: Driving Visual Geometry Transformer
arXiv:2512.16919v2 Announce Type: replace-cross Abstract: Perceiving and reconstructing 3D scene geometry from visual inputs is crucial for autonomous driving. However, there still lacks a driving-targeted dense geometry perception model that can adapt to different scenarios and camera configurations. To bridge this gap, we propose a Driving Visual Geometry Transformer (DVGT), which reconstructs a global dense 3D point map from a sequence of unposed multi-view visual inputs. We first extract visual features for each image using a DINO backbone, and employ alternating intra-view local attention, cross-view spatial attention, and cross-frame temporal attention to infer geometric relations across images. We then use multiple heads to decode a global point map in the ego coordinate of the first frame and the ego poses for each frame. Unlike conventional methods that rely on precise camera parameters, DVGT is free of explicit 3D geometric priors, enabling flexible processing of arbitrary camera configurations. DVGT directly predicts metric-scaled geometry from image sequences, eliminating the need for post-alignment with external sensors. Trained on a large mixture of driving datasets including nuScenes, OpenScene, Waymo, KITTI, and DDAD, DVGT significantly outperforms existing models on various scenarios. Code is available at https://github.com/wzzheng/DVGT.
Abstract:Perceiving and reconstructing 3D scene geometry from visual inputs is crucial for autonomous driving. However, there still lacks a driving-targeted dense geometry perception model that can adapt to different scenarios and camera configurations. To bridge this gap, we propose a Driving Visual Geometry Transformer (DVGT), which reconstructs a global dense 3D point map from a sequence of unposed multi-view visual inputs. We first extract visual features for each image using a DINO backbone, and employ alternating intra-view local attention, cross-view spatial attention, and cross-frame temporal attention to infer geometric relations across images. We then use multiple heads to decode a global point map in the ego coordinate of the first frame and the ego poses for each frame. Unlike conventional methods that rely on precise camera parameters, DVGT is free of explicit 3D geometric priors, enabling flexible processing of arbitrary camera configurations. DVGT directly predicts metric-scaled geometry from image sequences, eliminating the need for post-alignment with external sensors. Trained on a large mixture of driving datasets including nuScenes, OpenScene, Waymo, KITTI, and DDAD, DVGT significantly outperforms existing models on various scenarios. Code is available at this https URL.
Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Robotics (cs.RO)
arXiv:2512.16919 [cs.CV]
(or arXiv:2512.16919v2 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2512.16919 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 288 words
Related coverage
GeoWAM: Visual Geometry World Action Models for Autonomous Driving
arXiv · Aug 25, 2026
Why do models task game?
AI Alignment Forum · Aug 6, 2026
Valerant: An Automatic Navigable Game Map Generator via Action-Conditioned World Model Exploration
arXiv · Sep 11, 2026
GeoWorldAD: Geometry World Action Model for Autonomous Driving
arXiv · Jul 21, 2026