Rethinking Driving World Model as Synthetic Data Generator for Perception Tasks

TL;DR

Dream4Drive introduces a synthetic data generation framework that enhances perception tasks in autonomous driving by creating diverse, multi-view, photorealistic videos, significantly improving corner case detection and perception model performance.

Contribution

The paper presents Dream4Drive, a novel synthetic data generator that improves downstream perception tasks by enabling scalable, multi-view, photorealistic video creation and introduces the DriveObj3D dataset for diverse 3D-aware video editing.

Findings

Synthetic data boosts perception model performance across training epochs.

Dream4Drive generates multi-view corner cases at scale.

The framework enhances perception in autonomous driving scenarios.

Abstract

Recent advancements in driving world models enable controllable generation of high-quality RGB videos or multimodal videos. Existing methods primarily focus on metrics related to generation quality and controllability. However, they often overlook the evaluation of downstream perception tasks, which are $\mathbf{really\ crucial}$ for the performance of autonomous driving. Existing methods usually leverage a training strategy that first pretrains on synthetic data and finetunes on real data, resulting in twice the epochs compared to the baseline (real data only). When we double the epochs in the baseline, the benefit of synthetic data becomes negligible. To thoroughly demonstrate the benefit of synthetic data, we introduce Dream4Drive, a novel synthetic data generation framework designed for enhancing the downstream perception tasks. Dream4Drive first decomposes the input video into several 3D-aware guidance maps and subsequently renders the 3D assets onto these guidance maps. Finally, the driving world model is fine-tuned to produce the edited, multi-view photorealistic videos, which can be used to train the downstream perception models. Dream4Drive enables unprecedented flexibility in generating multi-view corner cases at scale, significantly boosting corner case perception in autonomous driving. To facilitate future research, we also contribute a large-scale 3D asset dataset named DriveObj3D, covering the typical categories in driving scenarios and enabling diverse 3D-aware video editing. We conduct comprehensive experiments to show that Dream4Drive can effectively boost the performance of downstream perception models under various training epochs. Page: https://wm-research.github.io/Dream4Drive/ GitHub Link: https://github.com/wm-research/Dream4Drive