Currently, Chip resistors identification problems mainly rely on artificial ways in the actual production process of a variety of defects, types, direction, positive and negative recognition, which not only slow detection, and long-term high cost, high rate of false positives. The following three types of convolution neural networks which based on the idea of the AlexNet model, the GoogLeNet model and the ResNet model can solve this problem.The training parameters of the three types of convolution neural networks are reduced to 4M(million), So it reach a very high identification accuracy rate and identification speed.
Training data download link: http://pan.baidu.com/s/1i4SGOPv password:kmyw