notesum.ai

Published at November 25

Ca2-VDM: Efficient Autoregressive Video Diffusion Model with Causal Generation and Cache Sharing

cs.CV

Released Date: November 25, 2024

Authors: Kaifeng Gao¹, Jiaxin Shi², Hanwang Zhang³, Chunping Wang⁴, Jun Xiao¹, Long Chen⁵

Aff.: ¹Zhejiang University; ²Xmax.AI; ³Nanyang Technological University; ⁴Finvolution Group; ⁵The Hong Kong University of Science and Technology

Arxiv: http://arxiv.org/abs/2411.16375v1

Refer to caption

Method	Res.	FVD
MCVD [41]	$64^{2}$	1143
VDT [21]	$64^{2}$	225.7
DIGAN^∗ [48]	$128^{2}$	577
TATS [8]	$128^{2}$	420
VideoFusion [23]	$128^{2}$	220
LVDM^∗ [14]	$256^{2}$	372
PVDM [49]	$256^{2}$	343.6
Latte [24]	$256^{2}$	333.6
Ca2-VDM	$256^{2}$	184.5