Enhancing Object Detection in Advanced Driver Assistance Systems Using Synthetic Weather ConditionsMUJIĆ, EMIR (Avtor) Perš, Janez (Mentor) Štern, Darko (Komentor) object detectiongenerative modelsGANQS-AttnADASADSsynthetic weather conditionsautonomous vehiclesEnsuring reliable object detection under adverse weather conditions, such as rain and snow, remains a significant challenge for visual sensing in the automotive world. These conditions can severely impact the performance of sensors, leading to reduced accuracy in object detection and an increased risk of accidents. This thesis focuses on enhancing object detection capabilities in ADAS/ADS by incorporating synthetic weather condition images into the training process in order to improve object detection in those conditions. Utilizing generative models we simulate the effects of adversarial weather on driving images in both city and highway environments. The primary objective is to create a comprehensive dataset of synthetic weather images, which, when integrated into the training pipeline for object detection, can improve the robustness of object detection models under adverse weather conditions. Our findings demonstrate that integrating synthetic weather images into the training process significantly enhances the performance of object detection models, making them more resilient to adverse weather conditions. This research not only addresses a critical safety concern in the automotive industry but also paves the way for future advancements in the field of autonomous driving.20242024-07-26 13:05:04Magistrsko delo/naloga159825VisID: 62726COBISS_ID: 203268867sl