基于卷积神经网络的图像识别教学实验

Image Recognition Teaching Experiment Based on Convolutional Neural Network

图像识别是"图像处理"教学中的重要内容。本文在Linux环境下使用i Torch notebook可视化界面利用卷积神经网络实现mnist手写数字体的准确识别,并详细介绍卷积神经网络的原理,给出直观的实验结果。教学实践表明,通过具有应用性和趣味性的实验可以提高学生的积极性,加深对课程理论的认知,培养其分析问题和解决问题的能力。

Image recognition is the important content of Image Processing course. In this paper,under the environment of Linux,we realize the accurate handwritten digit recognition from the data of mnist by using a visual interface called itorch notebook,introduce the principle of the convolutional neural network in detail,and give an intuitive experimental results. The teaching experience shows that such kind of experiment which embodies interestingness and practical application has the advantage of increasing the students＇ interest in the course,deepen the cognition of curriculum theory and cultivate their ability to analyze and solve problems.