arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1807.08024cs.CV
收藏

用反事实生成解释图像分类器

Explaining Image Classifiers by Counterfactual Generation

University of Toronto(多伦多大学);Vector Institute(向量研究所)

Chun-Hao Chang, Elliot Creager, Anna Goldenberg, David Duvenaud

摘要:

当图像分类器进行预测时,图像的哪些部分是相关的,为什么?我们可以将这个问题重新措辞为:如果分类器没有看到图像的哪些部分,它们最会改变其决策?产生一个答案需要边缘化的图像,可以看到,但没有。我们可以通过在图像的其余部分上调节生成模型来对合理的图像填充进行采样。然后,我们进行优化,以找到填充后最能改变分类器决策的图像区域。我们的方法与ad-hoc填充方法(例如模糊或注入噪声)形成对比,这些方法生成远离数据分布的输入,并忽略图像不同部分之间的信息关系。我们的方法产生更紧凑和相关的显着性图,与以前的方法相比,工件更少。

英文摘要:

When an image classifier makes a prediction, which parts of the image are relevant and why? We can rephrase this question to ask: which parts of the image, if they were not seen by the classifier, would most change its decision? Producing an answer requires marginalizing over images that could have been seen but weren't. We can sample plausible image in-fills by conditioning a generative model on the rest of the image. We then optimize to find the image regions that most change the classifier's decision after in-fill. Our approach contrasts with ad-hoc in-filling approaches, such as blurring or injecting noise, which generate inputs far from the data distribution, and ignore informative relationships between different parts of the image. Our method produces more compact and relevant saliency maps, with fewer artifacts compared to previous methods.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑