This document presents a thesis on using YOLO v5 for real-time object detection of potholes, speed breakers, and vehicles. It discusses the objectives, methodology, and implementation of training a YOLO v5 model. The methodology section outlines the steps for preparing the dataset, environment setup, model training, inference on test images, and result visualization. The results section shows various performance metrics and detected objects on test images. It concludes the proposed method provides a preliminary solution for road object detection to help road maintenance agencies and drivers.