Development of a device for sorting apples by variety using convolutional neural network technologies

The work is aimed at developing a convolutional neural network for automated sorting of apples by size, weight, and visual defects. Studies were carried out using an experimental installation, which was used to collect and markup the data array. The neural network was written in Python using the Tensorflow and Keras libraries. The created model exhibited an accuracy of 96.88 % in recognition results on delayed data. A sorter device model was designed to implement the trained algorithm in the sorting production cycle.

Authors: A. E. Yablokov, S. A. Potapov

Direction: Informatics, Computer Technologies And Control

Keywords: computer vision, neural networks, object recognition, classification, sorting

---

View full article