This project aims to design and implement a vision-based navigation system for an arc welding robot, leveraging deep learning techniques for automated welding point detection and integration. The system utilizes the YOLO (You Only Look Once) object detection algorithm to identify and localize relevant features such as weld seams or joint edges in real-time video streams. The development pipeline begins with data acquisition, where annotated images of weld joints are collected under varying lighting and environmental conditions. This dataset is then used to train a YOLO-based convolutional neural network (CNN) to perform precise object recognition. A Raspberry Pi serves as the central processing unit, interfacing with a camera module and a time-of-flight (ToF) or ultrasonic sensor to collect both visual and spatial data. These inputs are synchronized to extract accurate three-dimensional welding coordinates. The coordinates are then formatted and transmitted via a pre-programmed interface to a Yaskawa robotic controller, which interprets the data to guide the welding arm’s motion with high precision. The entire object detection system underwent rigorous validation through a series of tests that measure positional accuracy, detection reliability, and system responsiveness. Error detection algorithms are employed to identify and correct anomalies during the yolo model training phase, ensuring the robustness of the system. This research contributes to the advancement of intelligent robotic welding by reducing human intervention, improving weld consistency, and optimizing process efficiency. The integration of computer vision and real-time control enhances the reliability and repeatability of arc welding tasks in industrial manufacturing environments.