Aiming at the problems that the management performance of the existing stencil storage system is not strong and the detection efficiency of the storage stencil is low,a stencil storage system using deep learning method for pre-detection of the stencil is proposed.Firstly,an automatic system for stencil storage based on CAN bus is designed,and then the stencil data set is estab-lished by data enhancement method,and finally the ResNet18 model and MobileNetV2 model are used to train based on transfer learning to achieve pre-detection of stencil storage.Experimental results show that the stencil storage system can realize the automat-ic management of the stencil.At the same time,the ResNet18 model is applied to the pre-detection of stencil storage with an accura-cy rate of 98.45%,which can replace the work of manual selection of image features.
关键词
深度学习/钢网/迁移学习/卷积神经网络
Key words
deep learning/stencil/transfer learning/convolutional neural networks
维普
万方数据