@MASTERSTHESIS{pgi2020028,
	author = "P. Konstantinou",
	supervisor = "M. Goodfellow",
	title = "Real time Emotion Detection",
	school = "Department of Computer and Information Sciences, University of Strathclyde",
	year = "2019",
	abstract = "Computer vision is the subfield of artificial intelligence that deals with how computers can gain a high-level understanding from images. Examples of this include face detection and recognition systems, line tracking systems that are currently implemented on self-driving cars and authentication of documents such as passports or drivers{\^a}€™ licenses. The area of interest of this project revolves around the field of emotion recognition.  

This dissertation thesis outlines the development of a Web-Application that can run natively on a Raspberry pi with the main functionality of detecting emotions in an input video stream. Different solutions to this problem are compared and the most lightweight one is implemented, demonstrating that the modern minicomputers are capable of carrying out tasks as complex as this one. The emotion detection module of the system is comprised of a face detection and extraction algorithm in series with a face recognition algorithm trained to recognise emotions. During the development two face detection algorithms and three face recognition algorithms are compared, and the best performing algorithms are utilised.  

Results show that the system that has been developed can recognise emotions in an input video stream in real time, but when ran natively on the Raspberry Pi the limitations of the processing power impact the performance of the system. Emotion recognition works for pre-recorded video clips, but in real time the frame rate at which emotion detection happens is reduced.",
}