Explanation of Convolutional Neural Network for Automotive Wire Harness Using Gradient-Weighted Class Activation Mapping

This paper describes a methodology for explaining a convolutional neural network (CNN) for automotive wire harness using a gradient-weighted class activation mapping (Grad-CAM). Previous studies have achieved the judgement of the automotive wire harness using a machine learning (ML) technique. However, due to the black-box nature of ML-based methods, there is a lack of transparency, leading to a lack of trust in experimental results and hindering further research progress. In this paper, we propose how to obtain an activation map by applying Grad-CAM to explain the CNN model and get a more intuitive understanding of the mechanism behind harness diagnosis.