基于TensorFlow手写数字识别模型改进

Implementation of Handwritten Numbers Recognition Based on TensorFlow

设计手写数字识别模型,在TensorFlow框架上使用Softmax回归算法实现手写数字图像的识别,并且在此模型基础上添加神经网络改进模型。实验表明,使用TensorFlow能够快速实现手写图像识别,改进模型后实验准确率可从87.8%提升到99.2%。

In this paper,we designed a handwritten digital recognition model and used the Softmax regression algorithm to realize the recognition of handwritten digital images on the TensorFlow framework,and added a neural network improvement model based on this model.Experiments showed that using TensorFlow can quickly realize handwritten image recognition,and the experimental accuracy can be improved from 87.8%to 99.2%.