DiVE: DiT-based Video Generation with Enhanced Control

TL;DR

This paper introduces DiVE, a novel DiT-based framework for generating high-quality, temporally and multi-view consistent videos in autonomous driving scenarios, with precise control matching bird's-eye view layouts.

Contribution

It is the first DiT-based method specifically designed for multi-view, controllable video generation with cross-view consistency mechanisms.

Findings

Effective in producing long, controllable videos in challenging corner cases.

Outperforms existing methods in qualitative assessments on nuScenes dataset.

Ensures cross-view consistency through a novel spatial view-inflated attention mechanism.

Abstract

Generating high-fidelity, temporally consistent videos in autonomous driving scenarios faces a significant challenge, e.g. problematic maneuvers in corner cases. Despite recent video generation works are proposed to tackcle the mentioned problem, i.e. models built on top of Diffusion Transformers (DiT), works are still missing which are targeted on exploring the potential for multi-view videos generation scenarios. Noticeably, we propose the first DiT-based framework specifically designed for generating temporally and multi-view consistent videos which precisely match the given bird's-eye view layouts control. Specifically, the proposed framework leverages a parameter-free spatial view-inflated attention mechanism to guarantee the cross-view consistency, where joint cross-attention modules and ControlNet-Transformer are integrated to further improve the precision of control. To demonstrate our advantages, we extensively investigate the qualitative comparisons on nuScenes dataset, particularly in some most challenging corner cases. In summary, the effectiveness of our proposed method in producing long, controllable, and highly consistent videos under difficult conditions is proven to be effective.