<![CDATA[Oranges have provided the benefits of vitamin C to the human body, so it is necessary to cultivate local orange fruit varieties using the implementation of computer vision technology. With the optimization of accuracy results using the CNN method, one of which is a combination of YoloV9c and VGG-16 can be realized on local oranges to overcome the problem of inaccuracy in the inefficiency of the classification process influenced by human visual subjectivity so as not to produce inconsistency in local orange fruit defect detection. Optimization was carried out to obtain the best accuracy results of 97% in this study, compared to the accuracy results using the YoloV9c method alone of 74% or the VGG-16 method alone of 59%. The accuracy optimization used a frame rate dataset of 2,221 images which were divided into 1,555 training data, 444 testing data, and 222 validation data images with a percentage of sorting of 70% for training data, 20% for testing data, and 10% for validation data. With this research, it provides new insights and knowledge to combine the YoloV9c method with VGG-16, where VGG-16 is used in the data pre-processing stage using feature extraction and fine-tuning with a batch size of 32 while YoloV9c is used to classify the results of local orange fruit quality detection with 100 epochs. With the combination of the CNN algorithm, it can increase the accuracy value of the detection results.]]>
Copyrights © 2025
