Autonomous self-driven vehicles are a rapidly evolving research field in the automotive industry. This paper focuses on developing a custom dataset and training a model for detecting auto rickshaw using camera. The objective is to improve the safety and efficiency of autonomous navigation systems in urban environments, particularly in countries where auto rickshaw are commonly used for transportation, such as India, Bangladesh, Thailand, Sri Lanka, Indonesia, the Philippines, and parts of Africa. For this purpose, 5000 images of AR were utilized to train the YOLO algorithm for accurate labelling and to detect mixed traffic conditions. The real-time performance of the algorithm is evaluated using precision, recall, and Mean Average Precision (MAP) metrics. The goal is to contribute to the advancement of reliable and efficient autonomous navigation systems even in complex traffic conditions, enabling the widespread adoption of autonomous technology worldwide.