Εφαρμογή τεχνικών μηχανικής μάθησης για την ανάλυση τηλεδιασκέψεων σε πραγματικό χρόνο
View/
Keywords
Ζωντανή βιντεομετάδοση ; Ανάλυση συναισθημάτων ; OpenCV ; TensorFlow ; Caffe ; Javascript ; MQTT ; Python ; Μηχανική μάθησηAbstract
The present master thesis examines the idea and implementation of an application for live
teleconferencing, through the web browser, with live video and audio. These were made possible
through the groundbreaking WebRTC protocol which broadened the horizons of web browsers’
developers, since there was no analogous before. In addition, over the last few years the interest
in machine learning applications has increased and leaps ahead have been made in terms of
the technology and its applications. As it evolves, new ideas and implementations are proposed,
to find out its feasibility and its limits. In the present thesis, were studied and developed, the
creation of Convoluted Neural Networks as well as the use of ready, pre-trained models on the
video conferencing. The purpose of machine learning is to show the video conference participant,
the emotions of the participants in live time. Because the application is designed to run on web
browsers, technologies and implementation strategy, were crucial for an attractive result. To this
end, from the research done, specific tools were used, such as OpenVidu to manage users and
video conferencing, OpenCV to recognize faces in videos, and Tensοrflow, Caffe to recognize
emotions. The implementation was done mainly in Javascript programming language, using the
React library, Python programming language for the training of machine learning of models and
their analysis in real time. In addition, experiments were performed on different devices to find
the feasibility and needs of the application, as well as different topologies and infrastructure, to
integrate it into future scenarios, such as distributed systems and cloud computing.
Postgraduate Studies Programme
Πληροφοριακά Συστήματα και ΥπηρεσίεςDepartment
Σχολή Τεχνολογιών Πληροφορικής και Επικοινωνιών. Τμήμα Ψηφιακών ΣυστημάτωνNumber of pages
97Language
GreekCollections
Except where otherwise noted, this item's license is described as
Attribution-NonCommercial-NoDerivatives 4.0 Διεθνές
Attribution-NonCommercial-NoDerivatives 4.0 Διεθνές