magistrsko delo
Abstract
Mikro obrazni izrazi so kratke in subtilne obrazne mimike, ki jih ne znamo kontrolirati z živčnim sistemom. Posledično pojavitev takšnih mimik lahko med drugim kaže na zakrivanje iskrenih čustev. Analiza mikro obraznih izrazov najde uporabno vrednost predvsem v aplikacijah znotraj javnega varstva in klinične medicine. Raziskave in razvoj sistemov za prepoznavo in klasifikacijo mikro obraznih izrazov se osredotočajo na avtomatsko, algoritemsko prepoznavo, saj so takšni izrazi težki za prepoznavo s prostim očesom in večkrat ostanejo neopaženi. V tem magistrskem delu naredim pregled nekaterih metod globokega učenja za klasifikacijo mikro obraznih izrazov v enega od osnovnih čustev (pozitivno, negativno, presenečeno, drugo) in njihovih uspešnosti na dosegljivih podatkovnih množicah, ki vsebujejo video vzorce z nespontanimi mikro obraznimi izrazi.
Keywords
strojno učenje;globoko učenje;nevronske mreže;mikro obrazni izrazi;strojni vid;
Data
Language: |
Slovenian |
Year of publishing: |
2020 |
Typology: |
2.09 - Master's Thesis |
Organization: |
UL FMF - Faculty of Mathematics and Physics |
Publisher: |
[M. Garafolj] |
UDC: |
004.85 |
COBISS: |
58177027
|
Views: |
1725 |
Downloads: |
149 |
Average score: |
0 (0 votes) |
Metadata: |
|
Other data
Secondary language: |
English |
Secondary title: |
Micro-expression recognition using deep learning methods |
Secondary abstract: |
Micro-expressions are short and subtle facial expressions, that we do not control with our nervous system. Among other things can occurrence of such micro-expressions indicate an attempt at hiding the real emotion. Algorithmic analysis of micro-expressions finds its value in the fields of public safety and clinical medicine. Research and development in micro-expression analysis focuses on algorithmic approaches, since it is borderline impossible for the naked eye to spot micro-expressions. In this work I do an overview of some deep learning methods for the classification of micro-expressions into one of the basic emotions (positive, negative, surprise, other) and report on their success at doing so, on various data sets. |
Secondary keywords: |
copulas;probability of default;pairwise copula construction;vine copulas;Monte Carlo methods; |
Type (COBISS): |
Master's thesis/paper |
Study programme: |
0 |
Embargo end date (OpenAIRE): |
1970-01-01 |
Thesis comment: |
Univ. v Ljubljani, Fak. za matematiko in fiziko, Oddelek za matematiko, Finančna matematika - 2. stopnja |
Pages: |
60 str. |
ID: |
12418893 |