UTAR Institutional Repository

Face reidentification system to track factory visitors using OpenVINO

Wong, Yiek Heng (2020) Face reidentification system to track factory visitors using OpenVINO. Final Year Project, UTAR.

[img]
Preview
PDF
Download (4032Kb) | Preview

    Abstract

    This paper introduces a facial recognition-based visitor tracking and reidentification system using OpenVINO. The system can track people based on their faces to track their identities. OpenVINO make CNN deep learning inference on the edge. It supports heterogeneous execution across computer vision accelerators, decrease the time to market through function of libraries which optimized calls for OpenCV and OpenVX and pre-optimized kernels. Every book that I’ve referred to in the past few years has the author always mention that deep learning requires a lot of computational power to run on. That is the reason why we need GPU. GPU are usually used to train CNN models and include tons of calculations to make predictions. Company NVIDIA uses CUDA and cuDNN for deep learning, providing the best GPU and best software support. However, GPUs are relative extremely expensive which typically cost 2-3 times that of CPU compute instances. Hence, you may not run you deep learning model inferencing on inexpensive devices. For example, you won’t apply an expensive GPU that costs thousands of ringgits to an IoT devices such as (Raspberry PI or Intel UP Squared board). Fortunately, developer may abstract away this difficulty by implement OpenVINO onto IoT devices and make it use CPU to inference on real time. In addition, as the datasets grow larger, traditional neural network prediction methods will no longer be accurate and fast enough. OpenVINO affect the performance of inference, OpenVINO optimizes multiple calls in the traditional computer vision algorithm implemented in OpenCV, and performs specific optimizations for deep learning inference. Developer may get the benefit when using OpenCV with OpenVINO. In this project, the development environment, running environment and the playground for running OpenVINO will be setup. At the end, two system will be developed which are “Face Registration System” and “Face Reidentification System”. These two systems will apply to separate devices and make them run concurrently, so that the face reidentification process will be able run on embedded board without using any expensive component.

    Item Type: Final Year Project / Dissertation / Thesis (Final Year Project)
    Subjects: T Technology > T Technology (General)
    T Technology > TA Engineering (General). Civil engineering (General)
    Divisions: Faculty of Information and Communication Technology > Bachelor of Information Technology (Hons) Computer Engineering
    Depositing User: ML Main Library
    Date Deposited: 06 Jan 2021 15:05
    Last Modified: 06 Jan 2021 15:05
    URI: http://eprints.utar.edu.my/id/eprint/3833

    Actions (login required)

    View Item