While solving many practical problems,the methods based on convolutional neural network(CNN)have shown better performance than traditional machine learning methods.However,as to a specific problem and data set,an efficient CNN model can only be designed based on knowing domain knowledge completely.Usually it will take plenty of computing resource and time to finish the designing procedure.This paper proposes a new automatic machine learning method(named rdpsoCNN)for fabric defect detection,the random drift particle swarm optimization algorithm(RDPSO)is used to automatically find an optimal CNN structure of one-class deep support vector data description.The experimental results show that the proposed rdpsoCNN can find good CNN structures that achieve better detection performance compared to the state-of-art designs.
关键词
自动机器学习/卷积神经网络/单分类深度支持向量数据描述/随机漂移粒子群优化算法/瑕疵检测
Key words
automatic machine learning/convolutional neural network/one-class deep support vector data description/ran-dom drift particle swarm optimization/defect detection
维普
万方数据