Fruits Quality Detection using Deep Learning Models: A Meta- Analysis
M. Pyingkodi, K. Thenmozhi, M. Karthikeyan, K. Chitra, N.R. Wilfred Blessing, S. K. Pravin Kumar
Abstract
Deep learning has progressed significantly in recent years, making it an ideal choice for image processing applications. Recognition and classification of fruits using deep learning are considered as the most promising techniques for commercial and agricultural applications. Despite this, the researchers are still having difficulty in distinguishing fruits due to their similar colour, shape, and size. This research study intends to address some of the existing challenges by developing a novel framework for fruit identification and classification. Deep learning models such as U-Net, VGG19, and ResNet are used in this work to train and test these models. Here, two sets of data were used: pre-processed and augmented dataset. ResNet performs well with 98% accuracy on the initial set of data, but it consumes more time to compute. As a result, the deep learning models could recognize and classify 122 different types of fruits in real time. However, image augmentation will have no effect on the models' performance.