Analýza rozhodování detektorů objectů pro autonomní automobily

Abstract

The primary objective of this thesis is to explore the potential of employing Explainable AI (XAI) methods to gain insights into the decision-making processes of state-of-the-art object detectors for sequential data, with a focus on applications in autonomous driving. The research begins by reviewing historical and modern object detection approaches and identifying suitable XAI methods that are able to interpret decisions made by deep neural networks on image data, particularly those that highlight relevant parts of the input image. Subsequently, an appropriate dataset relevant to autonomous driving, containing sequential data, is identified. A suitable object detection model is selected and integrated with the chosen XAI method to explain the model's decisions on the dataset. Finally, the thesis involves a quantitative and qualitative analysis of the XAI outputs for the model's decisions, aiming to uncover patterns and dependencies within the explanations.The primary objective of this thesis is to explore the potential of employing Explainable AI (XAI) methods to gain insights into the decision-making processes of state-of-the-art object detectors for sequential data, with a focus on applications in autonomous driving. The research begins by reviewing historical and modern object detection approaches and identifying suitable XAI methods that are able to interpret decisions made by deep neural networks on image data, particularly those that highlight relevant parts of the input image. Subsequently, an appropriate dataset relevant to autonomous driving, containing sequential data, is identified. A suitable object detection model is selected and integrated with the chosen XAI method to explain the model's decisions on the dataset. Finally, the thesis involves a quantitative and qualitative analysis of the XAI outputs for the model's decisions, aiming to uncover patterns and dependencies within the explanations.