Explainable AI-Based Semantic Object Detection for Autonomous Vehicles

Abstract

A goal is to facilitate the recognition and segmentation of the route driven by autonomous vehicles through the use of machine learning (ML) models. The pixel-wise road detection task, the semantic segmentation architectures underwent training and comparison. By using XAI, the authors are able to interpret and read the predictions generated by these abstract models. They generated arguments for the recommended segmentation model for autonomous vehicle road detection using a range of XAI approaches. Supervised learning is enabled by KNN, Decision Tree, and Random Forest, which are the current algorithms used for comparison. On the other hand, the newly built K means clustering function best when paired for image processing since they are good with images. To display the results of computations for the evaluation parameters of each algorithm, including accuracy, sensitivity, recall and precision, tables, and the necessary features from the evaluation matrices are utilised. The k-means clustering system for explainable AI-based semantic object detection in automated cars achieves 94.58% accuracy on both train and test sets.