Robustness of transfer learning to image degradation

作者：

Highlights：

• Transfer learning algorithm is very robust in image classification.

• Transfer learning algorithm is insensitive to sample size beyond certain size number.

• Transfer learning algorithm is robust in presence of image cropping, image rotation, noise corruption, poor image resolution.

摘要

•Transfer learning algorithm is very robust in image classification.•Transfer learning algorithm is insensitive to sample size beyond certain size number.•Transfer learning algorithm is robust in presence of image cropping, image rotation, noise corruption, poor image resolution.

论文关键词：Image classification,Transfer learning,Convolutional neural network,Learning robustness,Image degradation

论文评审过程：Received 4 January 2021, Revised 18 July 2021, Accepted 4 September 2021, Available online 10 September 2021, Version of Record 22 September 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.115877