MADrive: Memory-Augmented Driving Scene Modeling

TL;DR

MADrive introduces a memory-augmented scene reconstruction framework for autonomous driving that replaces observed vehicles with similar 3D assets, enabling photorealistic scene alterations and improved scene modeling.

Contribution

It presents MADrive, a novel framework that incorporates external memory for vehicle replacement in scene reconstruction, enhancing realism and flexibility in autonomous driving scene modeling.

Findings

Enables photorealistic synthesis of altered driving scenes.

Replaces observed vehicles with similar 3D assets from a large-scale memory bank.

Demonstrates improved scene realism and flexibility in experiments.

Abstract

Recent advances in scene reconstruction have pushed toward highly realistic modeling of autonomous driving (AD) environments using 3D Gaussian splatting. However, the resulting reconstructions remain closely tied to the original observations and struggle to support photorealistic synthesis of significantly altered or novel driving scenarios. This work introduces MADrive, a memory-augmented reconstruction framework designed to extend the capabilities of existing scene reconstruction methods by replacing observed vehicles with visually similar 3D assets retrieved from a large-scale external memory bank. Specifically, we release MAD-Cars, a curated dataset of ${\sim}70$K 360{\deg} car videos captured in the wild and present a retrieval module that finds the most similar car instances in the memory bank, reconstructs the corresponding 3D assets from video, and integrates them into the target scene through orientation alignment and relighting. The resulting replacements provide complete multi-view representations of vehicles in the scene, enabling photorealistic synthesis of substantially altered configurations, as demonstrated in our experiments. Project page: https://yandex-research.github.io/madrive/