SAR-BagNet: An Ante-hoc Interpretable Recognition Model Based on Deep Network for SAR Image

Convolutional neural networks (CNNs) have been widely used in SAR image recognition and achieved high accuracy on some public datasets. However, due to the opacity of decision-making mechanism, reliability credibility CNNs are insufficient at present, which hinders their application important fields such as recognition. In recent years, various interpretable network structures proposed discern relationship between a CNN’s decision regions. Unfortunately, most based optical images, poor performance for them cannot accurately explain parts classification decisions. Based above problems, this study, we present SAR-BagNet, is novel framework images. SAR-BagNet can provide clear heatmap that reflect impact each part final decision. Except good interpretability, also has achieve 98.25% test accuracy.