基于深度学习的小包外观图像分类识别

Packet Appearance Image Classification and Recognition Based on Deep Learning

为了提高对小包外观图像的分类识别准确率和测试精度。设计了一种基于深度学习的视觉几何组网算法结构来加强对图像分类识别的准确度。使用视觉几何组网对图像分类识别的准确度能够达到96%以上,而传统浅层和深层卷积神经网络对图像分类识别的准确度只能够达到76%。使用视觉几何组网对小包外观图像进行分类识别准确率高,测试精度也有所提升。利用VGGNet-11与VGGNet-13分别对输入图片进行特征提取,并在最后对提取的特征进行耦合,以此来提高图像分类的准确性。

In order to improve the classification,recognition and test accuracy of the appearance image of the packet.A visual geometry group net algorithm structure based on deep learning is designed to enhance the accuracy of image classification and recognition.The accuracy of image classification and recognition using visual geometry group nets can reach 96% or higher,while the values of the accuracy of traditional shallow,deep convolutional neural networks for image classification and recognition can only reach 76%.The visual geometry group net is used to classify and recognize small packet appearance images achieving high accuracy and test accuracy.VGGNet-11 and VGGNet-13 are used to extract features from input images respectively and couple the extracted features at the end to improve the accuracy of image classification.