What do adversarial images tell us about human vision?
What do adversarial images tell us about human vision?
复制标题
DOI：
10.7554/elife.55978
复制
发表时间：
2020-09-02
期刊：
影响因子：
7.7
通讯作者：
Bowers, Jeffrey S.
中科院分区：
文献类型：
--
作者：
Dujmovic, Marin;Malhotra, Gaurav;Bowers, Jeffrey S.
深度卷积神经网络（DCNN）经常被描述为人类和灵长类动物视觉的最佳模型。对这一说法的一个明显挑战是存在欺骗DCNN但人类无法解释的对抗图像。然而，最近的研究表明，人类和DCNN在如何解释这些看似无意义的图像方面可能存在相似之处。我们重新分析了一篇备受瞩目的论文中的数据，并进行了五个实验，控制了这些图像的生成和选择的不同方式。我们发现，人类-DCNN的一致性比以前报道的要弱得多，而且这种弱一致性取决于对抗图像的选择和实验的设计。事实上，我们发现有一些众所周知的方法可以生成人类与DCNN不一致的图像。我们的结论是，对抗性图像仍然对使用DCNN作为人类视觉模型的理论家提出了挑战。
Deep convolutional neural networks (DCNNs) are frequently described as the best current models of human and primate vision. An obvious challenge to this claim is the existence of adversarial images that fool DCNNs but are uninterpretable to humans. However, recent research has suggested that there may be similarities in how humans and DCNNs interpret these seemingly nonsense images. We reanalysed data from a high-profile paper and conducted five experiments controlling for different ways in which these images can be generated and selected. We show human-DCNN agreement is much weaker and more variable than previously reported, and that the weak agreement is contingent on the choice of adversarial images and the design of the experiment. Indeed, we find there are well-known methods of generating images for which humans show no agreement with DCNNs. We conclude that adversarial images still pose a challenge to theorists using DCNNs as models of human vision.