Human Emotion Recognition in Images Using Machine Learning

DJINDJIC, PREDRAG (author), Jaklič, Aleš (mentor)

Identifying human facial expressions (FER or Face expression Recognition) is a complex and very important challenge in the spectre of social communication that researches and scientists have been tackling over the last few decades. One of the most obvious and simple ways for humans to express their feelings without speaking to each other are facial expressions. Facial expressions are crucial components for non-verbal communication between people. In contrast with the most FER systems found in research papers that only identify a small number of facial expressions, my thesis provides an automatic system for emotion detection on images that is capable of recognizing eight basic human expressions - neutral, angry, happy, sad, fear, disgust and contempt. I also describe and explain other FER approaches in this thesis as well as their contributions. We compare systems by a simple measure - number of correctly detected expressions on images.

• Language: English
• Keywords: machine learning, computer vision, face recognition
• Work type: Bachelor thesis/paper
• Typology: 2.11 - Undergraduate Thesis
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2022
• PID: 20.500.12556/RUL-140242
• COBISS.SI-ID: 121627907
• Publication date in RUL: 13.09.2022

Secondary language

• Language: Slovenian
• Title: Prepoznavanje čustev ljudi iz slik s strojnim učenjem

Abstract

Prepoznavanje čustev ljudi je zapleten in zelo pomemben izziv v spektru družbenega komuniciranja, s katerim se raziskovalci in znanstveniki ukvarjajo zadnjih nekaj desetletij. Eden izmed najbolj očitnih in preprostih načinov, da ljudje izrazijo svoja čustva brez pogovarjanja drug z drugim, so sami izrazi obraza (čustev). Obrazna mimika je ključna sestavina neverbalne komunikacije med ljudmi. V nasprotju z večino sistemov FER, ki jih najdemo v raziskovalnih člankih, ki identificirajo le majhno število obraznih izrazov, moj diplomski seminar zagotavlja samodejni sistem za zaznavanje čustev na slikah ki je sposoben prepoznati osem osnovnih človeških izrazov - nevtralen, jezen, vesel, žalosten, strah, gnus in prezir. V diplomskem delu opisujem in razlagam tudi druge pristope FER ter njihove prispevke. Sisteme primerjamo s preprostim merilom - številom pravilno zaznanih izrazov na slikah.

• Keywords: masinsko ucenje, raˇcunalniˇski vid, prepoznavanje obraza