Automated Detection and Classification of Returnable Packaging Based on YOLOV4 Algorithm
This article describes the implementation of the You Only Look Once (YOLO) detection algorithm for the detection of returnable packaging. The method of creating an original dataset and creating an augmented dataset is shown. The model was evaluated using mean Average Precision (mAP), F1<inline-fo...
Main Authors: | , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2022-11-01
|
Series: | Applied Sciences |
Subjects: | |
Online Access: | https://www.mdpi.com/2076-3417/12/21/11131 |
_version_ | 1797469182571315200 |
---|---|
author | Matko Glučina Sandi Baressi Šegota Nikola Anđelić Zlatan Car |
author_facet | Matko Glučina Sandi Baressi Šegota Nikola Anđelić Zlatan Car |
author_sort | Matko Glučina |
collection | DOAJ |
description | This article describes the implementation of the You Only Look Once (YOLO) detection algorithm for the detection of returnable packaging. The method of creating an original dataset and creating an augmented dataset is shown. The model was evaluated using mean Average Precision (mAP), F1<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><msub><mrow></mrow><mi>score</mi></msub></semantics></math></inline-formula>, Precision, Recall, Average Intersection over Union (Average IoU) score, and Average Loss. The training was conducted in four cycles, i.e., 6000, 8000, 10,000, and 20,000 max batches with three different activation functions Mish, ReLU, and Linear (used in 6000 and 8000 max batches). The influence train/test dataset ratio was also investigated. The conducted investigation showed that variation of hyperparameters (activation function and max batch sizes) have a significant influence on detection and classification accuracy with the best results obtained in the case of YOLO version 4 (YOLOV4) with the Mish activation function and max batch size of 20,000 that achieved the highest mAP of 99.96% and lowest average error of 0.3643. |
first_indexed | 2024-03-09T19:16:50Z |
format | Article |
id | doaj.art-21e44490867e4da28d5db9fecfa903c9 |
institution | Directory Open Access Journal |
issn | 2076-3417 |
language | English |
last_indexed | 2024-03-09T19:16:50Z |
publishDate | 2022-11-01 |
publisher | MDPI AG |
record_format | Article |
series | Applied Sciences |
spelling | doaj.art-21e44490867e4da28d5db9fecfa903c92023-11-24T03:38:31ZengMDPI AGApplied Sciences2076-34172022-11-0112211113110.3390/app122111131Automated Detection and Classification of Returnable Packaging Based on YOLOV4 AlgorithmMatko Glučina0Sandi Baressi Šegota1Nikola Anđelić2Zlatan Car3University of Rijeka, Trg Braće Mažuranića 10, 51000 Rijeka, CroatiaFaculty of Engineering, University of Rijeka, Vukovarska 58, 51000 Rijeka, CroatiaFaculty of Engineering, University of Rijeka, Vukovarska 58, 51000 Rijeka, CroatiaFaculty of Engineering, University of Rijeka, Vukovarska 58, 51000 Rijeka, CroatiaThis article describes the implementation of the You Only Look Once (YOLO) detection algorithm for the detection of returnable packaging. The method of creating an original dataset and creating an augmented dataset is shown. The model was evaluated using mean Average Precision (mAP), F1<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><msub><mrow></mrow><mi>score</mi></msub></semantics></math></inline-formula>, Precision, Recall, Average Intersection over Union (Average IoU) score, and Average Loss. The training was conducted in four cycles, i.e., 6000, 8000, 10,000, and 20,000 max batches with three different activation functions Mish, ReLU, and Linear (used in 6000 and 8000 max batches). The influence train/test dataset ratio was also investigated. The conducted investigation showed that variation of hyperparameters (activation function and max batch sizes) have a significant influence on detection and classification accuracy with the best results obtained in the case of YOLO version 4 (YOLOV4) with the Mish activation function and max batch size of 20,000 that achieved the highest mAP of 99.96% and lowest average error of 0.3643.https://www.mdpi.com/2076-3417/12/21/11131artificial intelligence algorithmsautomated systemconvolutional neural networkcomputer visionYOLOV4 |
spellingShingle | Matko Glučina Sandi Baressi Šegota Nikola Anđelić Zlatan Car Automated Detection and Classification of Returnable Packaging Based on YOLOV4 Algorithm Applied Sciences artificial intelligence algorithms automated system convolutional neural network computer vision YOLOV4 |
title | Automated Detection and Classification of Returnable Packaging Based on YOLOV4 Algorithm |
title_full | Automated Detection and Classification of Returnable Packaging Based on YOLOV4 Algorithm |
title_fullStr | Automated Detection and Classification of Returnable Packaging Based on YOLOV4 Algorithm |
title_full_unstemmed | Automated Detection and Classification of Returnable Packaging Based on YOLOV4 Algorithm |
title_short | Automated Detection and Classification of Returnable Packaging Based on YOLOV4 Algorithm |
title_sort | automated detection and classification of returnable packaging based on yolov4 algorithm |
topic | artificial intelligence algorithms automated system convolutional neural network computer vision YOLOV4 |
url | https://www.mdpi.com/2076-3417/12/21/11131 |
work_keys_str_mv | AT matkoglucina automateddetectionandclassificationofreturnablepackagingbasedonyolov4algorithm AT sandibaressisegota automateddetectionandclassificationofreturnablepackagingbasedonyolov4algorithm AT nikolaanđelic automateddetectionandclassificationofreturnablepackagingbasedonyolov4algorithm AT zlatancar automateddetectionandclassificationofreturnablepackagingbasedonyolov4algorithm |