BriefGPT.xyz
May, 2017
人类和深度学习在视觉扭曲下识别性能的研究与比较
A Study and Comparison of Human and Deep Learning Recognition Performance Under Visual Distortions
HTML
PDF
Samuel Dodge, Lina Karam
TL;DR
在图像质量失真的影响下，深度神经网络的表现远不及人类，但两者的错误率存在着较少的相关性，表明图像的内部表现在网络和人类眼中存在差异。这些与人类视觉表现的比较有助于指导未来更具鲁棒性的深度神经网络的发展。
Abstract
deep neural networks
(DNNs) achieve excellent performance on standard classification tasks. However, under
image quality distortions
such as blur and noise,
→