PreWorld introduces a two-stage semi-supervised training paradigm that achieves state-of-the-art 3D occupancy prediction and competitive 4D forecasting and planning on nuScenes using a combination of 2D and 3D supervision.
Lift, splat, shoot: Encoding images from arbitrary camera rigs by implicitly unprojecting to 3d
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Semi-Supervised Vision-Centric 3D Occupancy World Model for Autonomous Driving
PreWorld introduces a two-stage semi-supervised training paradigm that achieves state-of-the-art 3D occupancy prediction and competitive 4D forecasting and planning on nuScenes using a combination of 2D and 3D supervision.