基于Keras的图像分类模型的研究与应用

Research and Application of Image Classification Model Based on Keras

为提高图像识别的准确率,降低模型的损失,以及简化模型构建过程,文章提出一种基于Keras框架的图像分类神经网络模型。通过改进传统神经网络模型中损失函数以及增加DropOut过拟合技术,利用谷歌TensorFlow平台的Keras高级接口进行图像识别模型的搭建和训练,最终将模型应用于Fashion-MNIST多分类数据集,实验结果表明提出的基于Keras图像分类模型极大地简化了模型的复杂度,减少了模型过拟合现象的发生,并提高了图像分类的准确率。

In order to improve the accuracy of image recognition,reduce model loss,and simplify the model construction process,this article proposes an image classification neural network model based on the Keras framework.By improving the loss function in traditional neural network models and adding DropOut overfitting technology,the image recognition model was built and trained using the Keras advanced interface of Google TensorFlow platform.Finally,the model was applied to the Fashion MINIST multi classification dataset.The experimental results showed that the proposed Keras based image classification model greatly simplified the complexity of the model and reduced the occurrence of overfitting,And it has improved the accuracy of image classification.