A feature understanding method for explanation of image classification by convolutional neural networks

Convolutional Neural Networks (CNNs) are frequently used for tasks related to visual content understanding like classification and segmentation, primarily due to the state-of-the-art architectures that have been ubiquitously available. Though they show promising performances, fundamentally there are still questions about the safety and the trustworthiness of these solutions. In this chapter we present an improvement to the Feature Understanding Method (FEM) called the Modified FEM to explain the decisions of Deep CNNs trained for image classification tasks. The method belongs to the family of the so-called “white-box” methods. It is based on the analysis of the features in the last convolution layer of the network with further backpropagation to identify the image pixels which contributed to the decision the most. The method explains the decision of a trained network for a specific sample image. In the chapter this method has been applied for the explanation of the network trained for the classification of chest X-ray images for the recognition of COVID-19 disease.