{"type":"ScholarlyArticle","title":"Visual Implicit Geometry Transformer for Autonomous Driving","authors":[{"name":"Arsenii Shirokov","slug":"arsenii-shirokov","given":"Arsenii","family":"Shirokov","affiliation":""},{"name":"Mikhail Kuznetsov","slug":"mikhail-kuznetsov","given":"Mikhail","family":"Kuznetsov","affiliation":""},{"name":"Danila Stepochkin","slug":"danila-stepochkin","given":"Danila","family":"Stepochkin","affiliation":""},{"name":"Egor Evdokimov","slug":"egor-evdokimov","given":"Egor","family":"Evdokimov","affiliation":""},{"name":"Daniil Glazkov","slug":"daniil-glazkov","given":"Daniil","family":"Glazkov","affiliation":""},{"name":"Nikolay Patakin","slug":"nikolay-patakin","given":"Nikolay","family":"Patakin","affiliation":""},{"name":"Anton Konushin","slug":"anton-konushin","given":"Anton","family":"Konushin","affiliation":""},{"name":"Dmitry Senushkin","slug":"dmitry-senushkin","given":"Dmitry","family":"Senushkin","affiliation":""}],"abstract":"We introduce the Visual Implicit Geometry Transformer (ViGT), an autonomous driving geometric model that estimates continuous 3D occupancy fields from surround-view camera rigs. ViGT represents a step towards foundational geometric models for autonomous driving, prioritizing scalability, architectural simplicity, and generalization across diverse sensor configurations. Our approach achieves this through a calibration-free architecture, enabling a single model to adapt to different sensor setups. \r\n    Unlike general-purpose geometric foundational models that focus on pixel-aligned predictions, ViGT estimates a continuous 3D occupancy field in a bird’s-eye-view (BEV) addressing domain-specific requirements.\r\n    ViGT naturally infers geometry from multiple camera views into a single metric coordinate frame, providing a common representation for multiple geometric tasks.\r\n    Unlike most existing occupancy models, we adopt a self-supervised training procedure that leverages synchronized image-LiDAR pairs, eliminating the need for costly manual annotations.\r\n    We validate the scalability and generalizability of our approach by training our model on a mixture of five large-scale autonomous driving datasets (NuScenes, Waymo, NuPlan, ONCE, and Argoverse) and achieving state-of-the-art performance on the pointmap estimation task, with the best average rank across all evaluated baselines. \r\n    We further evaluate ViGT on the Occ3D-nuScenes benchmark, where ViGT achieves comparable performance with supervised methods.","doi":"10.24963/ijcai.2026/179","url":"https://ijcai.org/proceedings/2026/179","pdf":"https://ijcai.org/proceedings/2026/0179.pdf","markdown":"https://ijcai.org/proceedings/2026/179.md","bibtex":"https://ijcai.org/proceedings/2026/bibtex/179","volume":{"slug":"2026","title":"Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence","short_title":"IJCAI 2026","url":"https://ijcai.org/proceedings/2026/","year":2026,"location":"Bremen, Germany","dates":"15-21 August 2026","editors":"Diego Calvanese","isbn":"978-1-956792-09-6","publisher":"International Joint Conferences on Artificial Intelligence Organization","doi_prefix":"10.24963/ijcai.2026"},"section":"Computer Vision","track":"Main Track","number":179,"pages":{"first":1604,"last":1612},"keywords":["Computer Vision: 3D computer vision","Computer Vision: Scene analysis and understanding"],"videos":[],"references":[],"license_note":"Copyright International Joint Conferences on Artificial Intelligence."}