Mango Fruit Variety Classification Using Lightweight VGGNet Model

The Convolutional Neural Network (CNN) for mango fruit classification has yielded promising performance. However, because of inherent heavy-weight architectures, these approaches necessitate expensive training processes and more storage because they feed in an enormous amount of training parameters. In this paper, we have developed the lightweight VGGNet (LVGGNet) model to identify mango fruit varieties. The proposed LVGGNet model is trained and tested using a variety of datasets and validation methods. After performing some preprocessing on the data, we divide it into training and testing datasets. The convolutional layer, maxpool layer, and fully connected layer are the three layers of the CNN model that will be trained. We use the ReLU activation function to begin the model’s training and testing. In addition, a dataset containing 1600 images has been used in this endeavour. The empirical results demonstrate that the proposed LVGGNet model achieves 97.50% accuracy in identifying mango fruit variety.