FruitNet

There are many varieties of fruits that are available in the market, and they are different in shapes, colors, and textures. Choosing a quality fruit from many types of fruits is a challenging task to the common man and automated machines. In context, it is necessary to develop an efficient classification system. This problem has been generally solved by machine learning and deep learning models. A deep convolutional network has been adopted and compared with popular pre-trained CNN models like AlexNet. These prior approaches have resulted in high but still comparatively low accuracies and complex models. So in our paper, a dataset comprising of 6 categories of highly demanded fruits, namely orange, banana, pomegranate, lime, apple, guava, included in 3 folders of different quality of fruits: good, bad and mixed has been used. The dataset has 3 folders which are then further divided into 6 subfolders, totalling to 19,526 images which are then trained through image classification models like KNN, NN, CNN(with and without Regularization) and Transfer Learning. The best results were obtained by Transfer learning model.