09/02/2025
Advances in Computing and Engineering Journal
Volume4, Issue 2, 2024
https://apc.aast.edu/ojs/index.php/ACE/issue/view/77/showToc
Leveraging deep learning technology for enhancing printing press quality
Omotunde Alabi Muyiwa, B.H. Adejumo
DOI: https://dx.doi.org/10.21622/ACE.2024.04.2.951
Abstract
Machine learning technique usage for printing quality control is yet to be adopted in most printing press in Nigeria. However, deep learning technology can be used to improve printing quality. This study was designed to leverage deep learning technology for defect-detection in newspaper to improve printing quality. Six-hundred images of newspaper were loaded in pyCharm programmed environment for data exploration, cleaning, pre-processing, augmentation, while MATPLOT library analysed visual characteristics of loaded random sample image-dataset. A four-hundred newspaper-images were selected, which were divided into 320 (160-defective + 160 non-defective) for training, and 80 (40-defective + 40 non-defective) for validation and testing. The Convolutional Neural Networks (CNN), Support Vector Machines (SVM), Gaussian filters, Local Binary Patterns (LBP), pre-trained Visual Geometry Group sixteen (VGG16) models, Neural Network Search (NNS), and Deep Forest Models (DFM) were used for defect-detections. The CNN provided an acceptable image extraction feature for defect-detection, with validation accuracy of 66.7%. The machine learning ensemble classifiers of Gaussian filter+ LBP + SVM, CNN + SVM, simple CNN, transfer learning with VGG16, NNS, and gcForest gave training accuracy of 97.3, 71.5, 72.5, 81.3, 82.3, and 80%, respectively. These results demonstrated effectiveness of various machine learning techniques for defect-detection in newspaper images, which the Gaussian filter +LBP+SVM model achieved highest accuracy of 97.3%. The printing press can leverage on deep learning models to improve quality of the newspaper printing. The Gaussian filter+ LBP + SVM, CNN + SVM deep learning model should be adopted in printing press industry for high quality printing.Advances in Computing and Engineering Journal
Volume4, Issue 2, 2024
https://apc.aast.edu/ojs/index.php/ACE/issue/view/77/showToc
Leveraging deep learning technology for enhancing printing press quality
Omotunde Alabi Muyiwa, B.H. Adejumo
DOI: https://dx.doi.org/10.21622/ACE.2024.04.2.951
Abstract
Machine learning technique usage for printing quality control is yet to be adopted in most printing press in Nigeria. However, deep learning technology can be used to improve printing quality. This study was designed to leverage deep learning technology for defect-detection in newspaper to improve printing quality. Six-hundred images of newspaper were loaded in pyCharm programmed environment for data exploration, cleaning, pre-processing, augmentation, while MATPLOT library analysed visual characteristics of loaded random sample image-dataset. A four-hundred newspaper-images were selected, which were divided into 320 (160-defective + 160 non-defective) for training, and 80 (40-defective + 40 non-defective) for validation and testing. The Convolutional Neural Networks (CNN), Support Vector Machines (SVM), Gaussian filters, Local Binary Patterns (LBP), pre-trained Visual Geometry Group sixteen (VGG16) models, Neural Network Search (NNS), and Deep Forest Models (DFM) were used for defect-detections. The CNN provided an acceptable image extraction feature for defect-detection, with validation accuracy of 66.7%. The machine learning ensemble classifiers of Gaussian filter+ LBP + SVM, CNN + SVM, simple CNN, transfer learning with VGG16, NNS, and gcForest gave training accuracy of 97.3, 71.5, 72.5, 81.3, 82.3, and 80%, respectively. These results demonstrated effectiveness of various machine learning techniques for defect-detection in newspaper images, which the Gaussian filter +LBP+SVM model achieved highest accuracy of 97.3%. The printing press can leverage on deep learning models to improve quality of the newspaper printing. The Gaussian filter+ LBP + SVM, CNN + SVM deep learning model should be adopted in printing press industry for high quality printing.
