跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2611.19857 · v1

PanDWM: Panoramic 4D Driving World Models

首次提交 2026-11-15 15:45
版本更新 2026-11-15 15:45
雷达观测 2026-11-17 12:06
arXiv ↗PDF ↗

摘要

Driving world models usually generate videos from six pinhole cameras with separate optical centres and small overlaps, and must learn to keep these views consistent. We present PanDWM, which stitches them into one 360∘360^\circ panoramic band. Camera boundaries then fall inside one image, so keeping the views consistent means keeping the image continuous, and each pixel still records its camera ray. Conditioned on layouts, a causal student generates the band window by window in four steps; it is distilled from a causal teacher and trained on its own rollouts. A feed-forward decoder lifts each frame along these rays into 3D Gaussians, and static Gaussians from neighbouring frames complete them into 4D scenes. On nuScenes, scored in one stitched 360∘360^\circ view, PanDWM has lower FID and FVD than the multi-camera baselines at 88-46×46\times their speed, and keeps its 50-step teacher's FID while lowering FVD by 22-25%.

作者备注:Code: https://github.com/AIGeeksGroup/PanDWM, Website: https://aigeeksgroup.github.io/PanDWM

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

目前只收录了一个版本。收录下一版后,可以对照标题、作者、分类和摘要。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v1 · 2026-11-15 15:45 · 最新观测版本

PanDWM: Panoramic 4D Driving World Models

Driving world models usually generate videos from six pinhole cameras with separate optical centres and small overlaps, and must learn to keep these views consistent. We present PanDWM, which stitches them into one 360∘360^\circ panoramic band. Camera boundaries then fall inside one image, so keeping the views consistent means keeping the image continuous, and each pixel still records its camera ray. Conditioned on layouts, a causal student generates the band window by window in four steps; it is distilled from a causal teacher and trained on its own rollouts. A feed-forward decoder lifts each frame along these rays into 3D Gaussians, and static Gaussians from neighbouring frames complete them into 4D scenes. On nuScenes, scored in one stitched 360∘360^\circ view, PanDWM has lower FID and FVD than the multi-camera baselines at 88-46×46\times their speed, and keeps its 50-step teacher's FID while lowering FVD by 22-25%.

查看此版本解读 →查看此版本原文 ↗