Abstract:
Fruits with the same color, shape, and size are called identical fruits. There are a few tiny distinctions between the two, despite their identical appearance. In an effort to discriminate between identical fruits, a method has been created. The "big challenge" is to create object-detection technology on par with human ability. Deep learning can be a revolutionary solution to this problem. Deep learning makes it easier and quicker to make sense of massive datasets and draw conclusions from them. We used deep learning algorithms to detect the difference between identical fruits. In this research, we used YOLO v5 and the Faster R-CNN algorithm. We conducted the study using data from four categories: apple, pear, orange, and lemon. For the implementation of this process, we collect image data from fruit stores and label them according to our algorithms' requirements. We have augmented the image and made a data set of 1680 images for the methods to gain more accuracy. In our quest to identify fruits with comparable properties, we have explored both the Faster R-CNN and Yolo v5 algorithms. Yolo v5 obtained 97% accuracy compared to 91% for Faster R-CNN. Among the two algorithms, Yolo v5 produced the highest success rate, 97%. Our research will contribute to the advancement of automated agricultural technology. It is also useful in the supermarket to identify identical fruits. Most importantly it is a critical step toward automating processes such as automated harvesting and packaging etc.
