Real Time Fruit Recognition and Calories Estimation Using CNN and Model Deployment on Amazon's AWS

To carry out vigorous and effective fruit recognition in farming is difficult since there are a number of variations. Lately, various deep-learning approaches presented an outstanding result in numerous visual-guided agricultural usage. These approaches need labelling on training data, which is a tedious work. This study presented a customized five layers CNN architecture and various pretrained deep learning algorithms i.e., VGG-16, DenseNet-121 and InceptionV3are used to recognise the fruit and their calories for helping farmers, patients and peoples to estimate the fruit calories for their health and in farming. The model is trained and validated on integrated datasets which is downloaded from Kaggle and IEEE dataport having 10 classes of different fruits. After extensive experiments, highest accuracy obtained was 99.60% by five layers of CNN and proposed deep learning model has been deployed on Amazon AWS EC2 instance using WinSCP, PuTTY and PuTTYgen. Flask (Python Framework), HTML, CSS and JavaScript is used for website development and connected to its parent website which help users to recognize the fruit and their calories.