Vegetable Classification Using Pretrained Deep Learning Model for Accurate Identification

This study presents the development and evaluation of an automated vegetable classification system using the VGG16 convolutional neural network model. The increasing demand for efficient and accurate methods in agricultural practices underscores the need for such systems. A diverse dataset comprising images of various vegetables was employed, which was preprocessed by resizing to 224x224 pixels, normalizing pixel values, and applying data augmentation techniques. The VGG16 model, pre-trained on the ImageNet dataset, was fine-tuned with fully connected layers specifically designed for the vegetable classification task. The training process demonstrated significant improvements in both training and validation metrics over 20 epochs. The training loss decreased from 2.3256 in the first epoch to 0.214 by the twentieth epoch, while the training accuracy increased from 33.89% to 93.25%. Similarly, the validation loss showed a consistent downward trend, dropping from 0.1134 at the fourth epoch to 0.0429 at the twentieth epoch. The validation accuracy improved from 96.93% to 98.73% over the same period. These results highlight the VGG16 model's efficacy in accurately classifying vegetable images, with substantial improvements in learning and generalization capabilities. The consistent reduction in training and validation losses, alongside the steady increase in accuracies, underscores the model's robustness and reliability. This research provides a foundational framework for implementing deep learning techniques in automated vegetable classification, offering significant potential for enhancing efficiency, reducing labor costs, and improving quality control in agricultural and food production sectors. The high accuracy and reliability of the VGG16 model make it a promising tool for future advancements in AI-driven agricultural practices.