.png){kind=icon}
.png){kind=icon}
Return
Date Fruit Detection and Classification Based on Its Variety Using Deep Learning Technology
.png){kind=icon}
OA
AI
DOI：10.1109/ACCESS.2024.3433485.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
There are many types of date fruits, each with similar physical properties and slight differences in color, shape, and fleshiness. Due to the many attributes to consider, distinguishing between the different types of dates can be challenging and time consuming. Classifying and sorting the post-harvest fully ripe (Rutab) dates red into edible dates, is an essential stage of the date fruit production industries. Date production industries in Saudi Arabia still perform this process manually, which may result in reduced productivity and an increase in the date fruit waste. The purpose of this study was to use Deep Learning technology to detect and classify them by using the You Only Look Once (YOLO) algorithm. Different YOLO models (YOLOv5, YOLOv7, YOLOv8) have been trained using 1735 images with 9 types of date fruits. The accuracy of this study has been investigated using the metrics of Recall, Precision, and F1-score metrics. Based on the experimental results, YOLOv8 achieves a mean Recall of 0.99 % and Precision of 0.991% with an Intersection over Union (IoU) colorred in the range of [0-1], a mean Average Precision of 0.994%, a mean Average Precision of [0.5:0.95] and a mean Average Precision of [0.5:0.85]. Results show that the used model can accurately detect and classify date fruits based on their surface quality, enhancing the productivity as a result of multiple harvests due to uneven ripening of the fruit.
Keywords:
YOLO
Training
Accuracy
Measurement
Deep learning
Data models
Sorting
Artificial neural networks
Detection algorithms
Machine learning
Performance evaluation
Artificial neural network
deep learning
detection performance
image analysis
machine learning models