Asian Basic and Applied Research Journal

Object detection in dynamic situations is critical for applications such as traffic monitoring, surveillance, and public safety. Detecting and tracking many items is difficult due to changing lighting conditions, occlusions, and fast-moving objects. In this research, we provide a object recognition system based on the YOLOv8 deep learning model that can efficiently recognize automobiles and crowds in video streams.

The proposed system is built on a Convolutional Neural Net- work (CNN) architecture and is trained using a custom annotated dataset taken from RoboFlow. Data preprocessing techniques like normalization and augmentation are used to increase model generalization. The model is assessed using common performance criteria such as precision, recall, and F1-score.

Experimental results show that the proposed YOLOv8 model achieves an overall precision of 0.600, recall of 0.348, and an F1-score of 0.440. The system obtains a mAP@50 of 0.422 and mAP@50–95 of 0.169, demonstrating its capability to detect target objects in dynamic environments. These results indicate that the proposed approach can effectively support applications such as automated surveillance and intelligent traffic monitoring systems.