Přístupnostní navigace
E-application
Search Search Close
Bachelor's Thesis
Author of thesis: Bc. Václav Darvaš
Acad. year: 2025/2026
Supervisor: Ing. Ilona Janáková, Ph.D.
Reviewer: Ing. Peter Honec, Ph.D.
This bachelor’s thesis deals with the design and evaluation of automated optical inspection (AOI) systems and detecting microscopic defects on micro-lenses for Argotech a.s. The aim of the thesis is to compare three technologically different approaches to visual quality control: deterministic digital image processing methods (Top-Hat transformation), supervised classification using a convolutional neural network (ResNet18), and an unsupervised anomaly detection algorithm utilizing vector search (FAISS). As part of the solution, a hardware imaging setup was designed, and a real industrial dataset containing 1044 annotated images was created. The results showed that classical algorithmic approaches fail due to the visual variability of samples. A key contribution of this work for industrial practice is the implementation of a safety margin (the “grey zone”) into the artificial intelligence models, thereby transforming autonomous models into safe assistance systems. The most reliable solution was the unsupervised learning method (FAISS), which achieved a recall of 98.4% on defective samples while delegating one-third of the production to manual inspection. The work successfully served as a Proof of Concept, based on which the company decided to integrate commercial machine learning solutions.
Automated optical inspection, machine vision, image processing, artificial intelligence, convolutional neural networks, anomaly detection, deep learning, quality control, microoptics
Date of defence
16.06.2026
Result of the defence
Defended (thesis was successfully defended)
Grading
A
Process of defence
Student obhájil bakalářskou práci. Komise neměla žádné námitky k řešené práci. V průběhu odborné rozpravy student odpověděl na dotazy oponenta. Navazující otázka komise směřovala k použitým metodám strojového učení. Probíhala optimalizace velikosti neuronové sítě? Student odpověděl na položené otázky komise bez výhrad.
Language of thesis
Czech
Faculty
Fakulta elektrotechniky a komunikačních technologií
Department
Department of Control and Instrumentation
Study programme
Automation and Measurement (BPC-AMT)
Composition of Committee
doc. Ing. Jan Mikulka, Ph.D. (předseda) doc. Ing. Petr Blaha, Ph.D. (místopředseda) Ing. Ilona Janáková, Ph.D. (člen) Ing. Lukáš Kopečný, Ph.D. (člen) Ing. Tomáš Macho, Ph.D. (člen) Ing. Soňa Šedivá, Ph.D. (člen)
Supervisor’s reportIng. Ilona Janáková, Ph.D.
Grade proposed by supervisor: A
Reviewer’s reportIng. Peter Honec, Ph.D.
Grade proposed by reviewer: A
Responsibility: Mgr. et Mgr. Hana Odstrčilová