An Efficient Fruit Recognition Approach for Pregnant Women using Deep Convolutional Neural Networks

Abstract

Pregnancy is a period when women require extra care. It is necessary to make sure optimal nutritional level and a well-balanced diet. However, it is regrettable that a huge percentage of the Bangladeshi population is unaware of this issue. New mothers, in particular, who have no prior experience with this situation, suffer the most. During this time, an imbalance of nutrition and inappropriate fruit consumption might lead to major pregnancy complications. We used a deep learning approach called Convolutional neural network (CNN) to classify and recognize different types of fruit to solve this problem. Our eight fruit classes have been divided into two primary categories based on whether they are beneficial or avoidable. We gathered information about fruit consumption during pregnancy from gynecologists and various health articles and journals on the internet. People will learn what fruits they should eat and what fruits they should avoid from our eight fruit classes by reading our paper. We deployed two pre-trained CNN models, VGG19 and MobileNetV2, to complete this task. We received our result after running these two models on our collect dataset. We achieved a training accuracy of 94.06 % and a testing accuracy of 93.43 % for VGG19. In MobileNetV2, on the other hand, we achieved accuracy of 90.12 % for training and 89.05 % for testing. VGG19 requires less time than MobileNetV2 during the training phase.