.... Object detection plays a critical role in modern computer vision applications such as surveillance, autonomous systems, smart cities, and aerial monitoring. With the rapid growth of deep learning techniques, accurate and real-time detection of multiple objects from images and video streams has become increasingly feasible. However, deploying complex deep learning–based object detection models on resource-constrained platforms such as drones and edge devices remains a significant challenge due to limitations in computation power, memory, and energy consumption. To address these issues, this paper presents a scalable and efficient multi-object detection framework that combines lightweight processing at the data acquisition level with powerful deep learning–based inference on the server side. The proposed approach filters input frames using reference-based image comparison and transmits only significant frames for further processing, thereby reducing communication overhead and improving system efficiency. Advanced convolutional neural network–based object detection models are employed to accurately detect and classify multiple objects in real time. The architecture supports dynamic model updates, enabling the detection of new object categories without modifying the embedded system. Experimental analysis demonstrates that the proposed system achieves a balanced trade-off between detection accuracy, processing latency, and resource utilization, making it suitable for real-world applications such as aerial surveillance, traffic monitoring, and intelligent security systems. Index Terms Multi-Object Detection, Deep Learning, Computer Vision, Real-Time Detection, YOLO, Drone-Based Surveillance, Edge–Server Collaboration
Bediskar et al. (Thu,) studied this question.