SLS Powder bed defects

Published: 24 March 2021| Version 1 | DOI: 10.17632/2yzjmp52fw.1
Contributors:
,

Description

This is a real manufacturing dataset that contains images of the powder bed surface of a selective laser sintering system. The images in this dataset were used to monitor and document the quality of the printing process. The dataset consists of two folders. The data folder contains 8514 images of OK and DEF powder bed images. The data_balanced folder is subdivided into three subgroups, each with separate directories for two classes, OK and DEF. The training subgroup contains 1000 images in each class, the validation subgroup contains 500 images in each class and the test subgroup contains also 500 images in each class. The data_balanced folder thus consists of 4000 images that were copied from the data folder and then specially processed and classified into the subgroups and classes. The images for each subgroup were randomly selected from the data folder. Due to the lack of DEF images, some DEF images have been copied into the respective directories. It was ensured that no picture appears twice in the individual subgroups and subdirectories. A more detailed description of the dataset can be found in the associated article: https://doi.org/10.1016/j.addma.2021.101965

Files

Steps to reproduce

The steps to reproduce the data are described in the associated article: https://doi.org/10.1016/j.addma.2021.101965

Institutions

Universitat Rostock Fakultat fur Maschinenbau und Schiffstechnik

Categories

Machine Learning, Three Dimensional Printing, Quality Assurance, Selective Laser Sintering, Convolutional Neural Network

Licence