OneFixer improves 3D driving scene render quality and consistency fast
OneFixer: High-Quality and Consistent One-Step Autoregressive 3DGS Refinement for Driving Scenes
Computer Vision and Pattern RecognitionArtificial Intelligence
Summary
Simulating driving scenes in 3D can be tricky because each frame depends on the last, often leading to blurry or inconsistent video. The authors developed OneFixer, a model that fixes these visuals all at once while learning to avoid errors in future frames. It uses the model’s own predictions during training to better prepare for real use, improving both image quality and smoothness over time. This method also uses extra information like road lanes and moving cars to make the scene more accurate and realistic.
What this means in practice
- •For autonomous vehicle simulation teams: Improve visual quality and temporal smoothness of 3D driving simulations with faster one-step fixes that reduce collisions in closed-loop testing.
- •For 3d graphics developers: Use lane and agent information to enhance geometric detail and reliability in autoregressive 3D Gaussian splatting video renderers.
Authors
Boseong Jeon, Junhyeop Lee, Juhan Cha, Hayoung Kim
Abstract
Autoregressive video diffusion is a promising render-time fixer for 3D Gaussian Splatting (3DGS) in autonomous-driving simulation, but deployment demands high visual quality and temporal consistency at low latency. This is especially hard for one-step causal generation, where each imperfect prediction immediately becomes context for subsequent frames. Existing approaches stabilize rollouts through staged training with multiple modules and rollout-aware regularization, yet one-step quality still falls short of what deployment requires. We introduce OneFixer, a one-step autoregressive video-diffusion fixer trained in a single task-specific adaptation stage. Our key idea is a deployment-matched shared rollout: the model's own one-step predictions serve as the causal context for flow matching, exposing training to deployment-time errors, while the same rollout receives direct pixel-space perceptual supervision to preserve fine detail. Because the predictions optimized for current-frame quality are exactly those reused as future context, fidelity and autoregressive robustness are learned jointly, without bidirectional-to-causal conversion or teacher-student distillation. OneFixer further exploits cues that driving simulation readily provides, lane geometry and dynamic-agent states, to improve geometric fidelity. On Waymo and proprietary driving scenes with 900-frame rollouts, OneFixer achieves the lowest FVD, LPIPS, and DISTS among all baselines at one step, with temporal consistency matching or exceeding multi-stage DMD pipelines. Under identical backbone and conditioning, it matches a multi-stage DMD-with-Self-Forcing pipeline in under half the GPU-hours and keeps improving beyond its plateau. In closed-loop simulation with a driving policy, OneFixer reduces the collision rate by a third relative to raw 3DGS rendering. Project page: https://onefixer-web.vercel.app/