Σχολή Επιστημών και Τεχνολογίας της Πληροφορίας

Μόνιμο URI για αυτήν την κοινότηταhttps://pyxida.aueb.gr/handle/123456789/2

Η Σχολή Επιστημών και Τεχνολογίας της Πληροφορίας περιλαμβάνει τα Τμήματα: - Τμήμα Πληροφορικής - Τμήμα Στατιστικής

Περιήγηση

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Α Β Γ Δ Ε Ζ Η Θ Ι Κ Λ Μ Ν Ξ Ο Π Ρ Σ Τ Υ Φ Χ Ψ Ω

Τώρα δείχνει 1 - 1 από 1

Implementation of fault tolerant Big Data system for video / sensor analytics
(12/21/2018) Betchavas, Panagiotis-Ioannis; Kotidis, Yannis; Vassalos, Vasilios; Athens University of Economics and Business, Department of Informatics; Arkoumanis, Dinos
This master’s thesis defines the architecture for a fault tolerant Big Data system for video/sensor analytics along with an implementation of a running prototype. The goal is to build an on-premise system that is fault tolerant so it can provide continue correct performance of its specified tasks in presence of failure.To build such a system we will make use of an open source time series database called KairosDB which will run on top of a 2-node Scylla cluster with high replication factor to increase the fault tolerance of the developed system. Prometheus will also be included to the system for its read and write protocols, real-time monitoring and numerous adapters which can import and export data between Prometheus and various other databases like KairosDB. Finally, Grafana will allow us to query, visualize, alert on and understand the collected metrics no matter where they are stored.The video analysis is accomplished by two computer vision models; object detection and pose estimation. We will implement the YOLO model with Open CV library for object detection and the COCO model with OpenPose library for pose estimation using the Python Programming Language.In the first chapter, we have a brief overview of the master's thesis.In the second chapter, we introduce the different components of the system and describe the benefits that each of them provides to it.Next, we have the third chapter where the steps for the development of the system are presented. The focus of this chapter is the installation and configuration of ScyllaDB, KairosDB, Prometheus and Grafana so that they can communicate with each other.In the fourth chapter, we start with a brief introduction to object detection and pose estimation. Then, we specify the libraries and the models we used for object detection and pose estimation and we describe how the data move and processed from the webcam to KairosDB.Finally, the fifth chapter includes the conclusion and a brief representation of further improvements or future work on both the system itself and the implemented computer vision models.

Περιήγηση

Πλοήγηση Σχολή Επιστημών και Τεχνολογίας της Πληροφορίας ανά Επιβλέπων "Arkoumanis, Dinos"