Fruit Recognition and Classification Using Deep Learning (Case Study Date Plant)

EasyChair Preprint 7292

Abstract

Canvolutional neural network with a unique structure for feature extraction and classification is a new technology in the field of machine vision. Today, one of the challenges is to use deep learning in agriculture, harvest and quality-based packaging for domestic use as well as export industry . this study presents a method for discriminating and separating healthy dates, as well as predicting the ripening stage of dates.

The data set of this research was collected using a smart phone camera under uncontrolled conditions from shahani dates, which are specific to Iran. This data set includes images in four classes, three classes related to  khalal, rutab, tamar which are the stages of date maturation and one class related to defective dates. this research constructed from convolutional neural network methods, pre-trained networks such as VGG16 and Resnet. The Accuracy was measured by adding images obtained from Histogram of Oriented Gradient (HOG) to the model and using CLR call back technique The model was able to achieve an overall classification accuracy of 98%.

Keyphrases: Classification, Convolutional Neural Network, Histogram of Oriented Gradient, date fruit, deep learning, maturity stages

@booklet{EasyChair:7292,
  author    = {Maryam Talkhabi and Mahboobeh Shamsi and Majid Aghaei},
  title     = {Fruit Recognition and Classification Using Deep Learning (Case Study Date Plant)},
  howpublished = {EasyChair Preprint 7292},
  year      = {EasyChair, 2022}}