Road accidents, which are currently a big problem in Bangladesh, are largely caused by unsafe vehicles. Thousands of people are dying in Bangladesh due to road crushes. The huge number of unfit vehicles across the country has been increasing over the years, often leading to dangerous road accidents. Many lives are lost and thousands of people are injured in road accidents. According to the statistics, At least 5.42 lakh registered vehicles have been operating on the road without fitness clearance. Every year, 1.35 million people worldwide die in road accidents. In Bangladesh absence of proper monitoring of road transport authority, unfit vehicles can be operating on the road. To identify fit and unfit vehicles some major parts are checked on a vehicle such as the wheel system, braking system, steering system, body, tyres, etc. body parts are very important to identify the fitness status of the vehicle in the initial stage. Hence this paper proposes a vehicle fitness detection model by judging vehicle body parts. In this study only worked with bus which is major category of vehicle. Bus fitness detection using object detection techniques is a very challenging task because of the different colors, shapes, and sizes of vehicles. In this study deep learning-based YOLOv5 method is used for detecting unfit vehicles. YOLOv5 works based on CNN (Convolution Neural Network) architecture. It is a stable model of the YOLO family and has the ability to detect small objects. Data has been collected on the field. In this study Performed data annotation, preprocessing, augmentation, and data labeling for the different classes of vehicles. Divide the dataset into three parts for the purposes of testing, validation, and training. After training the model based on YOLOv5 it has tested on different images and real-time footage and achieved the mAP of 97.3% for all classes. Overall, the approach can be used for vehicle fitness monitoring purposes. It will also help to open new research scope for more improvement in the vehicle fitness detection field.