Real Time Object Detection using Pre trained deep learning models

The proposed system leverages pre‑trained deep learning architectures such as YOLOv8, SSD, and Faster R‑CNN to perform efficient multi‑object detection and classification. By utilizing advanced convolutional neural networks along with transfer learning strategies, the framework achieves real‑time detection with enhanced speed, accuracy, and computational efficiency. The adoption of pre‑trained models reduces training time, minimizes hardware requirements, and lowers dataset dependency, while simultaneously improving feature extraction and detection reliability. Experimental results demonstrated that the YOLOv8-based framework provides accurate and robust object detection performance even under challenging environmental conditions such as noise, partial occlusion, background complexity, and varying illumination levels. Furthermore, the proposed system supports deployment on embedded and edge computing devices, making it suitable for practical applications including intelligent surveillance, autonomous vehicles, robotics, industrial automation, healthcare monitoring, and smart IoT systems. Future research may focus on improving small object detection accuracy, integrating transformer-based detection architectures, reducing false-positive detections, and optimizing lightweight models for ultra-low-power embedded platforms. Additionally, the incorporation of advanced attention mechanisms and edge AI optimization techniques can further enhance real-time performance and scalability in next-generation intelligent vision systems.

[1] Y. Zhao, H. Liu, X. Wang, and L. Chen, “RT-DETR: Real-time detection transformer for efficient object detection,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 8, pp. 10215–10228, 2023. [2] Z. Chen, J. Huang, P. Li, and Y. Xu, “YOLO-MS: Multi-scale feature learning for real-time object detection,” Journal of Visual Communication and Image Representation, vol. 92, p. 103745, 2023. [3] S. More, R. Patil, and A. Kulkarni, “Comparative analysis of CNN, YOLO, and SSD for real-time object detection,” International Journal of Computer Vision and Robotics, vol. 14, no. 2, pp. 145–160, 2024. [4] H. Sun, Q. Zhang, and W. Liu, “D-YOLO: Attention-based feature fusion for robust object detection in adverse environments,” Pattern Recognition Letters, vol. 180, pp. 55–63, 2024. [5] M. Shobaki, F. Alenezi, and S. Rahman, “Comparative evaluation of YOLOv8, SSD, and Faster R-CNN using transfer learning techniques,” IEEE Access, vol. 13, pp. 22415–22432, 2025. [6] K. Ayachi, M. Ben Salem, and H. Trabelsi, “Transfer learning with YOLO variants for urban traffic object detection,” Expert Systems with Applications, vol. 250, p. 123456, 2025. [7] I. Ennaama, M. Rahmani, and S. El Fkihi, “Lightweight object detection using YOLOv7 and MobileNetV3 for edge devices,” Sensors, vol. 25, no. 4, p. 1887, 2025. [8] L. Zheng, T. Wang, and Y. Hu, “Mobile-EFSSD: Enhanced SSD with MobileNetV3 and feature pyramid networks,” Neurocomputing, vol. 612, pp. 128–141, 2025. [9] A. Aldubaikhi, M. Alqahtani, and R. Hassan, “Small object detection techniques in deep learning: A comprehensive review,” Artificial Intelligence Review, vol. 58, no. 3, pp. 1–35, 2025. [10] C. Edozie, P. Samuel, and A. Ibrahim, “A survey of deep learning-based object detection frameworks and applications,” Journal of Artificial Intelligence Research, vol. 78, pp. 301–340, 2025.