Strawberry Quality Classification Utilizing Convolutional Neural Network

Strawberry quality has been a crucial factor when it comes to consumer satisfaction. Having quality and cost-efficient strawberries would increase consumer satisfaction while increasing sales from the merchants’ perspective. As such, being able to classify strawberries into the desirable and undesirable categories would aid small businesses and consumers in determining whether the strawberries sold and bought are desirable based on key indicators such as shape and color. To address that, this study was conducted with the use of convolutional neural networks. The strawberry datasets used are a combination of a pre-classified dataset from another study and a dataset of images photographed solely for the purpose of this study. The images are classified as desirable and undesirable wherein 350 images of each set are used for training, 200 images for validation, and 100 images for testing. The generated CNN model is simulated using epochs=15 and batch size=8. The resulting CNN model has a training accuracy=98.41%, a validation accuracy=92.75%, and a testing accuracy=100% which makes the model efficient in classifying strawberries into the desirable and undesirable categories.