Machine learning approach in predicting GlutoPeak test parameters from image data with AutoML and transfer learning

Takehiro Murai; Yoshitaka Inoue; Assey Nambirige; George A Annor

doi:10.1016/j.heliyon.2023.e20522

Machine learning approach in predicting GlutoPeak test parameters from image data with AutoML and transfer learning

Heliyon. 2023 Sep 28;9(10):e20522. doi: 10.1016/j.heliyon.2023.e20522. eCollection 2023 Oct.

Authors

Takehiro Murai¹, Yoshitaka Inoue², Assey Nambirige¹, George A Annor¹

Affiliations

¹ Department of Food Science and Nutrition, College of Food, Agricultural and Natural Resource Sciences, University of Minnesota, 1334 Eckles Avenue, St. Paul, MN, 55108, USA.
² Department of Computer Science and Engineering, College of Science and Engineering, University of Minnesota, 200 Union St SE, Minneapolis, MN, 55455, USA.

Abstract

This study introduces a novel machine learning methodology for predicting GlutoPeak test parameters from image data, leveraging AutoKeras and transfer learning. The GlutoPeak test is a tool used in the baking industry to evaluate the properties of flour, based on its gluten strength and elasticity. Our research aimed to devise an efficient and cost-effective technique for quantifying the gluten properties of wheat varieties. We aimed to accomplish this by predicting the GlutoPeak test results with convolutional neural network (CNN) models, utilizing the benefits of transfer learning and AutoKeras. AutoKeras is a public code repository capable of automating neural architecture search and hyperparameter tuning. The ResNet101 model, when trained with the Adam optimizer, achieved the highest accuracy of 0.5765 in the 2-class prediction. Meanwhile, the ResNet101 model trained with the SGD optimizer reached the highest accuracy of 0.4362 in the 4-class prediction. The outcomes of this study illustrate the possibility in using machine learning and deep learning techniques for predicting GlutoPeak test parameters from image data. This offers a faster and more cost-effective approach for evaluating gluten quality in wheat varieties.

Keywords: AutoML; Convolutional neural network; Gluten; GlutoPeak test; Machine learning; Transfer learning.