Effect of Hyperparameter Tuning to Fruit Quality Detection Using Convolutional Neural Network

Malaysia's agricultural sector plays a crucial role in the nation's economy and food security. One of the most pressing issues is the difficulty in identifying and addressing fruit damage early in the production cycle. Current manual inspection methods are labor-intensive, time-consuming, and prone to human error. Convolutional Neural Network (CNN) is a useful tool that help in automatically learning and recognizing pattern in image and to determine the quality of the fruit in agriculture industry. This project is focusing on building CNN machine learning for image classification and detection of fruit quality. The CNN model is being trained with 3 different data sets which are apple, damage apple and potato. The model is trained for two-class-set (apple, damage_apple) and three-class-set (apple, damage_apple and potato) with 1000 and 1500 images consecutively. The experimental results in this study demonstrate that the hyperparameter adjustment of machine learning algorithms has varying effects on fruit quality detection.