notesum.ai

Published at November 29

ReconDreamer: Crafting World Models for Driving Scene Reconstruction via Online Restoration

cs.CV

cs.AI

cs.LG

cs.RO

Released Date: November 29, 2024

Authors: Chaojun Ni¹, Guosheng Zhao¹, Xiaofeng Wang¹, Zheng Zhu¹, Wenkang Qin¹, Guan Huang¹, Chen Liu², Yuyin Chen², Yida Wang², Xueyang Zhang², Yifei Zhan², Kun Zhan², Peng Jia², Xianpeng Lang², Xingang Wang³, Wenjun Mei⁴

Aff.: ¹GigaAI; ²Li Auto Inc.; ³CASIA; ⁴Peking University

Arxiv: http://arxiv.org/pdf/2411.19548v1

Method	Lane Shift @ 3m			Lane Shift @ 6m			Lane Change			Average
Method	NTA-IoU $\uparrow$	NTL-IoU $\uparrow$	FID $\downarrow$	NTA-IoU $\uparrow$	NTL-IoU $\uparrow$	FID $\downarrow$	NTA-IoU $\uparrow$	NTL-IoU $\uparrow$	FID $\downarrow$	NTA-IoU $\uparrow$	NTL-IoU $\uparrow$	FID $\downarrow$
PVG [8]	0.242	49.99	134.74	0.118	48.32	212.19	0.256	50.70	105.29	0.205	49.67	150.74
$\text{S}^{3}$ Gaussian [24]	0.059	48.24	234.54	0.014	47.69	311.59	0.175	49.05	124.90	0.083	48.33	223.68
Deformable-GS [62]	0.142	50.63	161.20	0.069	48.86	260.83	0.240	51.62	92.34	0.150	50.37	171.46
Street Gaussians [58]	0.498	53.19	130.75	0.374	49.27	213.04	0.496	56.03	86.46	0.456	52.83	143.42
ReconDreamer with	0.539	54.58	93.56	0.467	52.58	149.19	0.554	56.63	73.91	0.517	54.60	105.55
Street Gaussians	0.539	54.58	93.56	0.467	52.58	149.19	0.554	56.63	73.91	0.517	54.60	105.55