A New Deep Learning Application Based on Movidius NCS for Embedded Object Detection and Recognition
Year: 2018
Type: conference-paper
Abstract: Nowadays, real-time detection and recognition of objects is a vital task in image processing and computer vision. This study presents an embedded powerful technique for real-time object detection and recognition that runs at high frames per second (FPS) on an embedded platform with movidius neural compute stick (NCS). This can be done by applying a deep learning for computer vision. We recommended an object detection and recognition for real-time video by using deep learning technique and OpenCV libraries. It includes the single shot detector (SSD) algorithm with a MobileNet architecture that are trained with caffe framework. In this paper, Raspberry Pi 3 was utilized to implement this system. So, it helps to monitor and captures the frames and detect and recognize the objects. Also, we used movidius neural compute stick that can be utilized with the Raspberry Pi 3 to achieve high FPS. The proposed method applies a few enhancements such as default boxes, multi scale features and depthwise separable convolution. These enhancements permit the proposed system to get a high accuracy in detection and recognition of objects. Engineering.
Authors: Nashwan Adnan OTHMAN, Ilhan AYDIN
Institution: Fırat University
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2904939249
