Residual Neural Networks for Gemstone Recognition and Classification
<p>Xuhan Jin<sup>1</sup>, Tao Li<sup>2</sup></p> · Academic Journal of Computing & Information Science · 2022
For rapidly and accurately identifying and classifying different gemstones, a residual neural network-based gemstone classification and recognition model is presented using the image feature differences of 15 classes of gemstones. The gemstone image set is firstly established, then the image set is expanded using the data enhancement method, and afterwards the sample data set is obtained using the data cleaning method. The images are divided into training and test sets in the ratio of 8:1, and then the training set is trained using a residual neural network (ResNet 50). Finally, the correctness of the network is evaluated. The experimental results show that the average accuracy of this gem classification and identification algorithm reaches 93.46%, and basic engineering applications can be achieved.