notesum.ai

Published at December 3

cs.CR

Released Date: December 3, 2024

Authors: Qilong Wu¹, Varun Chandrasekaran¹

Aff.: ¹University of Illinois Urbana-Champaign

Method	Encoder	Decoder	$\ell$ (bits)	Training Dataset
HiDDeN [53]	5/8 Conv-BN-ReLU blocks and one Conv layer	Conv layers, average pooling, linear layer	20/30/64	DiffusionDB [45] or MidJourney [44]
MBRS [24]	Conv-BN-ReLU, ExpandNet, SENet blocks	Conv-BN-ReLU, SENet blocks, Conv layer, reshaping	64	DALLE-2 [23]
RivaGAN [49]	Conv3D-Tanh-BN3D blocks	Conv3D-Tanh-BN3D blocks	32	Hollywood 2 [31]
StegaStamp [43]	U-Net style architecture composed of Conv-ReLU blocks	Spatial transformer, Conv and dense layers, sigmoid	100	MIRFLICKR [21]