Software for the automatic identification of defective products using deep learning

Result code in IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F46747885%3A24210%2F22%3A00010935" target="_blank" >RIV/46747885:24210/22:00010935 - isvavai.cz</a>

Result on the web

<a href="https://github.com/petrsima/Batchelor-thesis" target="_blank" >https://github.com/petrsima/Batchelor-thesis</a>

DOI - Digital Object Identifier

—

Result language

angličtina

Original language name

Software for the automatic identification of defective products using deep learning

Original language description

This software utilizes machine learning techniques to detect anomalies in images of manufactured products. The software uses an autoencoder algorithm for detecting anomalies and defines the threshold for anomalous objects using kernel density estimation (KDE) and reconstruction loss. The software is designed to be adaptable to different types of products and can be customized to meet specific needs. With its powerful anomaly detection capabilities, this software can help identify faulty products, improving quality control and reducing waste.

Czech name

—

Czech description

—

Type

R - Software

CEP classification

—

OECD FORD branch

20206 - Computer hardware and architecture

Project

—

Continuities

I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Publication year

2022

Confidentiality

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Internal product ID

980b94a

Technical parameters

This software has the following technical parameters: - Input image size: 1024x1024 pixels in RGB format, but it can also work with smaller images. - Minimum number of images required for training: 200. - Programming language: Python, used to develop the application. - Machine learning library: Tensorflow, used for building and training the autoencoder model. - Anomaly detection techniques: This software uses the kernel density estimation (KDE) and reconstruction loss techniques available in the scikit-learn library (sklearn) to define the threshold for detecting anomalous objects in the images. - The software uses convolutional neural networks (CNNs) as the underlying architecture for the autoencoder. The model is trained using a dataset of images, and then it can be used to detect anomalies in new images. - The software‘s performance can be further optimized by adjusting parameters such as the number of hidden layers, the learning rate, and the activation functions used in the model. - The output of the software is a binary classification of the input images into normal or anomalous. - The software can be deployed on a variety of platforms and operating systems, and can be integrated into existing workflows and systems for quality control in manufacturing..

Economical parameters

As an open-source software, this anomaly detection tool can provide significant economic value to manufacturers by enabling them to identify faulty products early in the production process. By reducing the number of defective products that make it to the market, manufacturers can improve their brand reputation and reduce the costs associated with returns, repairs, and warranty claims. Additionally, the open-source nature of the software allows for customization and collaboration, enabling manufacturers to tailor the tool to their specific needs and benefit from the collective knowledge of the community. Overall, the use of this open-source anomaly detection tool can lead to increased efficiency, reduced waste, and improved profitability for manufacturers..

Owner IČO

46747885

Owner name

Technická univerzita v Liberci