Enhancing quality-based classification of perishable products: a convolutional neural network approach with statistical hyperparameter optimization

Ashish Kumar,Sunil Agrawal
DOI: https://doi.org/10.1007/s11042-024-19700-z
IF: 2.577
2024-06-21
Multimedia Tools and Applications
Abstract:India is among the leading producers of fresh fruits and vegetables in the world. Out of the total production, a significant portion is wasted annually due to multiple issues resulting from ineffective and inefficient postharvest supply chain processes. Quality is one of the important characteristics of the fruits and vegetables that is greatly affected by the perishability dynamics of the product and governs the buying tendency of the customer. Therefore, we propose an image processing model for the classification of perishable products based on their freshness that defines the overall quality of the product. The objective is to create an accurate and efficient Convolutional Neural Network (CNN) model for the classification of fruits and vegetables into fresh and rotten classes. The novelty of the research work is two-fold. First is the development of a novel five-layer CNN model for the classification task. Second is the application of statistical technique i.e., Design of Experiments (DoE) to optimize the hyperparameters of the CNN model. It resulted in obtaining an optimum accuracy of 99.58% on the test dataset. A comparison is also done between the developed and optimized CNN model with pre-trained models like AlexNet, VGG11, VGG13, VGG16, VGG19, ResNet18 and ResNet34 in terms of performance criterion i.e., accuracy and execution time. Also, comparison is done with transformer-based models to highlight the efficacy of our work. The DoE optimized model surpassed AlexNet, VGG11, VGG13, VGG16 and VGG19 in terms of chosen performance criterion while ResNet18 and ResNet34 model results were slightly better. Through the proposed work, quality control can be achieved by identifying and segregating fresh items from rotten ones leading to a reduction in the postharvest losses.
computer science, information systems, theory & methods,engineering, electrical & electronic, software engineering
What problem does this paper attempt to address?