Abstract In agricultural farming, defective fruits are the primary cause of global financial disasters. It has an impact on the reliability as well as the quality of the fruits. After harvest, quality inspection necessitates a lot of time and labour-intensive expertise. As a result, saving time and labour during harvest is made possible by automatically detecting fruit quality. With machine learning and image processing techniques, numerous algorithms have been developed to identify and classify fruit quality. A system incorporating Convolutional Neural Networks (CNN) and transfer learning methods have been created to advance the fruit categorization process. Two models are proposed to estimate fruit freshness. One customized CNN architecture is suggested by adjusting the network's parameters to fit the dataset. The second method uses the pre-trained VGG model and the transfer learning approach to determine the fruit's freshness. The suggested models can distinguish fresh and rotten fruit based on the input images. This study used 70 different kinds of fruit, including apples, bananas, oranges, and more. CNN collect features from input images of fruit, and then CNNs with specific categories are used to classify the input images. Because of the model's tailored design, when applied to a Kaggle dataset, the suggested model achieves a 99.39% accuracy on the training data and a 99.99% accuracy on the validation data. The model correctly classified 99.41% of the data. Transfer learning resulted in a 97.65% increase in classification accuracy, a 99.05% increase in training accuracy, and a 99.99% increase in validation accuracy. The results showed that the suggested model could distinguish between fresh and rotten fruit and applicable real-time farming applications.
Alan : Mühendislik
Dergi Türü : Uluslararası
Benzer Makaleler | Yazar | # |
---|
Makale | Yazar | # |
---|