Paired mini-batch training: A new deep network training for image forensics and steganalysis
作者：
Highlights：
•
• We explain why training convnets with the standard mini-batch is inefficient.
• We propose a new mini-batch, paired mini-batch, for image forensics and steganalysis.
• Training speed was faster when using the paired mini-batch.
• Test accuracy was higher when using the paired mini-batch.
• Proposed paired mini-batch can be applied to any convnets.
摘要
•We explain why training convnets with the standard mini-batch is inefficient.•We propose a new mini-batch, paired mini-batch, for image forensics and steganalysis.•Training speed was faster when using the paired mini-batch.•Test accuracy was higher when using the paired mini-batch.•Proposed paired mini-batch can be applied to any convnets.
论文关键词：Deep learning,Deep convolutional neural networks,Mini-batch,Image forensics,Steganalysis
论文评审过程：Received 30 November 2017, Revised 2 March 2018, Accepted 29 April 2018, Available online 15 June 2018, Version of Record 22 June 2018.
论文官网地址：https://doi.org/10.1016/j.image.2018.04.015
