YOLOv8 for Robust Traffic Object Instance Segmentation using Image Quality Assessment

Detecting objects in traffic plays a vital role across different domains, including autonomous driving, surveillance of traffic, and managing smart cities. The proposed work investigates the potential of deep learning models, specifically yolov8, for robust multi-spectral traffic object instance segmentation with a novel approach to image quality assessment. This study seeks to enhance object detection accuracy and performance in difficult traffic scenarios, including low light conditions, motion blur, and high noise levels. The results demonstrate the effectiveness of the hybrid image quality assessment approach in achieving robust object detection, showcasing a notable improvement in performance compared to baseline models and yielding promising results in comparison to state-of-the-art methods. The proposed framework contributes to the advancement of autonomous vehicle perception systems by offering improved accuracy and robustness in complex traffic environments.