Article
Multiclass Image Classification Using GANs and CNN Based
on Holes Drilled in Laminated Chipboard
Grzegorz Wieczorek
1,
* , Marcin Chlebus
2,
* , Janusz Gajda
2
, Katarzyna Chyrowicz
3
, Kamila Kontna
3
,
Michał Korycki
3
, Albina Jegorowa
4
and Michał Kruk
1
Citation: Wieczorek, G.; Chlebus, M.;
Gajda, J.; Chyrowicz, K.; Kontna, K.;
Korycki, M.; Jegorowa, A.; Kruk, M.
Multiclass Image Classification Using
GANs and CNN Based on Holes
Drilled in Laminated Chipboard.
Sensors 2021, 21, 8077. https://
doi.org/10.3390/s21238077
Academic Editors: Panagiotis E.
Pintelas, Sotiris Kotsiantis, Ioannis E.
Livieris and Anastasios Doulamis
Received: 28 September 2021
Accepted: 26 November 2021
Published: 2 December 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
1
Institute of Information Technology, Warsaw University of Life Sciences—SGGW, 02-787 Warsaw, Poland;
michal_kruk@sggw.edu.pl
2
Faculty of Economic Sciences, University of Warsaw, 00-927 Warsaw, Poland; jgajda@wne.uw.edu.pl
3
Data Juice Lab sp. z o.o., 00-503 Warsaw, Poland; k.chyrowicz@datajuicelab.com (K.C.);
k.kontna@datajuicelab.com (K.K.); m.korycki@datajuicelab.com (M.K.)
4
Institute of Wood Sciences and Furniture, Warsaw University of Life Sciences—SGGW,
02-787 Warsaw, Poland; albina_jegorowa@sggw.edu.pl
* Correspondence: grzegorz_wieczorek@sggw.edu.pl (G.W.); mchlebus@wne.uw.edu.pl (M.C.)
Abstract:
The multiclass prediction approach to the problem of recognizing the state of the drill by
classifying images of drilled holes into three classes is presented. Expert judgement was made on the
basis of the quality of the hole, by dividing the collected photographs into the classes: “very fine,”
“acceptable,” and “unacceptable.” The aim of the research was to create a model capable of identifying
different levels of quality of the holes, where the reduced quality would serve as a warning that the
drill is about to wear down. This could reduce the damage caused by a blunt tool. To perform this
task, real-world data were gathered, normalized, and scaled down, and additional instances were
created with the use of data-augmentation techniques, a self-developed transformation, and with
general adversarial networks. This approach also allowed us to achieve a slight rebalance of the
dataset, by creating higher numbers of images belonging to the less-represented classes. The datasets
generated were then fed into a series of convolutional neural networks, with different numbers of
convolution layers used, modelled to carry out the multiclass prediction. The performance of the
so-designed model was compared to predictions generated by Microsoft’s Custom Vision service,
trained on the same data, which was treated as the benchmark. Several trained models obtained by
adjusting the structure and hyperparameters of the model were able to provide better recognition of
less-represented classes than the benchmark.
Keywords: multi-class classification; laminated chipboard; GAN; CNN
1. Introduction
The quality of a drill and its impact on the quality of a final product, which was a piece
of furniture here, is of great importance in the production process. A drill that is not sharp
enough should be replaced in order to prevent it from damaging the products, which
would cause inconvenience and would generate costs to the producer.
The judgement of the state of a drill is not simple, and relying only on an expert’s eye
would be quite risky. A traditional approach to this problem is collecting and measuring
multiple signals produced by the drill, like the feed force, the cutting torque, the noise, the
vibration, or the acoustic emission and then estimating its quality based on these signals [
1
].
This approach gives acceptably accurate results, as it was shown in previous works [
2
–
4
],
but it requires the usage of multiple sensors. Many pre-processing operations have to
be performed on collected data, such as calculating a number of statistical parameters of
recorded signals or generating Fourier representations for specific feature selection [1].
In [
5
–
7
], it was shown that using only images of drilled holes and convolutional neural
networks (CNN) can give satisfying results, and it is a much simpler solution than that
Sensors 2021, 21, 8077. https://doi.org/10.3390/s21238077 https://www.mdpi.com/journal/sensors
