Published Paper Details:

The main objective of the project is to detect pedestrians using the YOLOv5 deep learning model and enhance pedestrian safety. The goal of this innovative pedestrian detection and safety method is to address the shortcomings of the existing system, improving detection accuracy and enhancing pedestrian safety. Leveraging advancements from previous YOLO versions, the YOLOv5 model offers increased speed and accuracy while providing a unified framework for model training, enabling real-time object detection, which is gaining substantial attention. Detecting pedestrians is particularly critical for autonomous vehicles, yet it poses significant challenges due to variations in age, gender, clothing, lighting, backgrounds, and occlusion among pedestrians. The paper introduces a novel approach to pedestrian detection using a publicly available image dataset. Creating accurate annotations for this dataset is challenging, given the presence of images with extreme lighting conditions, such as direct sunlight, and unreliable intrinsic features for training. Therefore, we propose a customized YOLOv5 model trained with different loss and regularization methods to enhance its base accuracy. Based on our findings, we present a solution for addressing real-time pedestrian detection issues. These techniques hold promise for various applications, including Autonomous Vehicles and Advanced Driver Assistance Systems (ADAS).