Object detection and recognition in realtime is the key task in many computer vision applications such as security surveillance, medical diagnosis, automated vehicle systems, etc. Now-a-days many deep learning techniques, especially convolutional neural networks (CNN) is widely used for real-time image detection and classification. The development of CNN models boosts the accuracy of object detection. However, the complex and data-intensive processing slows down the performance while implemented on hardware. This paper presents a low-powered, portable prototype on Xilinx PYNQ Z2 board with Movidius neural compute stick (NCS) that accelerates the object detection in real-time. Also, the proposed prototype utilized You Only Look Once (YOLO) approach for object detection. Frames per second (FPS), computation time and the probability of object recognition are the parameters considered to evaluate the performance of the proposed prototype and outperform the existing models.