Classification of Ripeness Fruits Using Convolutional Neural Network (CNN) With Multivariate Analysis by SGD in VGG16

The major backbone for a country economic growth is agriculture industry with high yield productivity and development in production. One of the essential process in fruit is estimating of fruit ripeness that influence the fruits quality as well as its marketing. Almost 30% to 35% have been made wasted while fruits harvesting due to lack of skilled workers. The evaluation of automatic fruit ripeness is executed though computer vision arrangement. Image processing has assisted in agricultural industry for improving the estimation of yield, detecting diseases, sorting of fruits and vegetables, irrigation as well as maturity grading. However, these challenges are handled in image processing with several hyperparameters that have livened up by conditional robust Convolutional Neural Network (CNN). The CNN help to obtain accurate classifying images with 16 layers (VGG16) with batch. Therefore, this paper focuses on Stochastic Gradient Descent (SGD) method which has used epoch concept over training dataset for computing the loss. It even update the parameters to identify as well as classify the images of rotten and ripen. The real time dataset is collected with RBG images consists of 6702 images involving three kinds of fruits namely oranges, papaya and banana with six classes such as three classes with the ripped fruits and another three classes with the rotten fruits. Moreover, the evaluation of CNN with VGG16 using SGD algorithm for accurate classification can be determined by parameters like accuracy and loss curve for training dataset as well as validation dataset with different epochs.