【摘要】 CNN卷积神经网络是图像识别和分类等领域的前沿研究方法。由于CNN模型训练效果与实际测试之间存在较大的差距,为提高自由手写数字的识别率,尝试使用TensorFlow搭构CNN网络模型,在完成MNIST数据集训练的基础上实现对自由手写数字的识别,并根据两种样本状态的差别和识别结果提出效果分析及改进方法,实验证明该改进方法获得明显效果。更多还原

【Abstract】 Convolution neural network is an advanced research method in the field of image recognition and classification. There is a big gap between the training effect and the actual test in the CNN model. To improve the recognition rate of free handwritten numeral, a CNN numeral model based on tensorflow is constructed to perform recognition by having trained on the MNIST datasets. The effect analysis and improvement method are put forward according to the differences of two samples’ status. Experiment shows the obviously improved result.更多还原