Publication Details

Intelligent Defect and Anomaly Detection System - A Study in Tofu

HRADIŠ Michal. Intelligent Defect and Anomaly Detection System - A Study in Tofu. Brno: Fingera s.r.o., 2015.
Czech title
Inteligentní detekce defektů a anomálií
Type
summary report
Language
english
Authors
Hradiš Michal, Ing., Ph.D. (DCGM FIT BUT)
Keywords

anomaly detection, defect detection, visual quality control, semi-supervised learning, active learning, convolutional neural networks

Abstract
The report proposes an intelligent system for detection of defects in images based on a number of computer vision and pattern recognition methods: classification, segmentation, anomaly detection, and classification support visualization. 
The components were tested on real-life data from a food packaging plant.
The studied methods have to be able to detect physical defects and anomalies of various types of products and they should be able to adapt to new types of products and defects automatically or by users supervision. The user should be able to guide the detection system by providing examples of defects, by correcting wrong detections of the automated system, or by similar actions.
Published
2015
Pages
1-35
Publisher
Fingera s.r.o.
Place
Brno, CZ
Back to top