Implementation of Model Pruning Methods for Resnet

The field of intelligent manufacturing is developing rapidly, with the number of network edge devices and the data generated growing rapidly. Technology of intelligent edge monitoring for surface defects has emerged, but the resources of edge devices are limited, making it difficult to deploy deep neural network models. This article proposes to use model pruning method to compress the network structure in response to this issue and implements the proposed method on ResNet network. The model reduces the number of parameters to 1/5 of the original while ensuring detection accuracy, and improves inference speed by 50%, ensuring the realtime detection of edge convolutional neural networks and improving the efficiency of varistor defect classification.