MNIST数据库
卷积神经网络
计算机科学
深度学习
人工智能
模式识别(心理学)
过程(计算)
特征提取
机器学习
集合(抽象数据类型)
程序设计语言
操作系统
作者
Baifang Niu,Hu Li,Wenxin Chen,Hongjun Han
标识
DOI:10.1109/icsp58490.2023.10248879
摘要
At present, the inspection of power plant equipment generally relies on labor, while the control of unsafe behaviors of personnel completely depends on the on-site supervision of safety supervisors. With the development of artificial intelligence, the intelligent inspection and fault recognition technology of power plants based on video AI algorithm have made breakthroughs, among which the core intelligent recognition and infrared image conversion are both based on picture recognition. This paper introduces the background of picture recognition and the development of deep learning. The paper studies the use of convolutional deep confidence network to identify MNIST, and shows the process of using this network to identify the experiment of handwritten number set, and the final recognition rate reached 99.2%. Finally, the accuracy of the most commonly used convolutional neural network is compared with the convolutional deep confidence network, and a control experiment is designed to study the influence of the learning rate and the iteration number on the recognition rate. Experiments show that convolutional deep confidence networks can be easily competent in identifying handwritten numbers.
科研通智能强力驱动
Strongly Powered by AbleSci AI