Facilitated machine learning for image-based fruit quality assessment

TL;DR

This paper introduces a simplified machine learning approach using pre-trained Vision Transformers for image-based fruit quality assessment, reducing data requirements and implementation complexity compared to traditional CNN methods.

Contribution

Proposes a machine learning procedure based on pre-trained Vision Transformers that is easier to implement and requires less training data than CNNs for fruit quality assessment.

Findings

Achieves comparable accuracy to CNNs within 1%

Requires three times fewer training samples for 90% accuracy

Effective for apple defect detection and banana ripeness estimation

Abstract

Image-based machine learning models can be used to make the sorting and grading of agricultural products more efficient. In many regions, implementing such systems can be difficult due to the lack of centralization and automation of postharvest supply chains. Stakeholders are often too small to specialize in machine learning, and large training data sets are unavailable. We propose a machine learning procedure for images based on pre-trained Vision Transformers. It is easier to implement than the current standard approach of training Convolutional Neural Networks (CNNs) as we do not (re-)train deep neural networks. We evaluate our approach based on two data sets for apple defect detection and banana ripeness estimation. Our model achieves a competitive classification accuracy equal to or less than one percent below the best-performing CNN. At the same time, it requires three times fewer training samples to achieve a 90% accuracy.