Identification of hickory nuts with different oxidation levels by integrating self-supervised and supervised learning

Abstract

The hickory (Carya cathayensis) nuts are considered as a traditional nut in Asia due to nutritional components such as phenols and steroids, amino acids and minerals, and especially high levels of unsaturated fatty acids. However, the edible quality of hickory nuts is rapidly deteriorated by oxidative rancidity. Deeper Masked autoencoders (DEEPMAE) with a unique structure for automatically extracting some features that could be scaleable from local to global for image classification, has been considered to be a state-of-the-art computer vision technique for grading tasks. This paper aims to present a novel and accurate method for grading hickory nuts with different oxidation levels. Owing to the use of self-supervised and supervised processes, this method is able to predict images of hickory nuts with different oxidation levels effectively, i.e., DEEPMAE can predict the oxidation level of nuts. The proposed DEEPMAE model was constructed from Vision Transformer (VIT) architecture which was followed by Masked autoencoders(MAE). This model was trained and tested on image datasets containing four classes, and the differences between these classes were mainly caused by varying levels of oxidation over time. The DEEPMAE model was able to achieve an overall classification accuracy of 96.14% on the validation set and 96.42% on the test set. The results on the suggested model demonstrated that the application of the DEEPMAE model might be a promising method for grading hickory nuts with different levels of oxidation.