Autonomous mobile robots are designed to navigate independently without human intervention. As the demand for these robots increases, new techniques and algorithms are continuously being developed. However, the use of ultrasonic sensors in vision-based obstacle avoidance systems has limitations. One significant disadvantage is their limited range compared to other sensor types, making it difficult to detect objects at longer distances or from wider angles, which may result in blind spots. Additionally, environmental conditions influence item identification accuracy because soft or absorbent surfaces can absorb ultrasonic waves. When many ultrasonic sensors operate simultaneously, there is a possibility of crosstalk or interference, which can result in inaccurate distance readings and obstacle detection. This research aims to develop an autonomous mobile robot capable of avoiding collisions using vision-based methods. The methodology includes capturing approximately 1609 images with a mobile robot-mounted camera, processing these images with the You Only Look Once (YOLO) algorithm for object detection, and converting the results into steering commands. Techniques such as labeling and data augmentation, including flipping and adjusting the brightness, were employed to enhance the dataset. The data was divided into training, validation, and testing sets with a 70:20:10 ratio split for classification. The integration of the YOLO algorithm achieved a detection success rate of approximately 93.3%. This study provides insights into the effectiveness of combining deep learning methods for real-time obstacle avoidance in mobile robotics, highlighting the importance of carefully considering the individual application and environment when selecting sensors for vision-based techniques.