Improving image recognition by hierarchical model and denoising
Fuqiang Chen, Yan Xiang Wu · 2015
In this study, a novel method for image recognition based on deep learning algorithm and image denoising is proposed. It is based on Bernoulli process factor analysis denoising method and tiled convolutional neural network. The images are first denoised using Bernoulli process factor analysis, and then the denoised images are transmitted to tiled convolutional neural network for robust feature extraction. Lastly, support vector machine is used for classification. The experiments implemented on the benchmark dataset CIFAR-10 shows the effectiveness of our proposed method, which performs better than our previously proposed method, CDAE-SVM (contractive denoising autoencoder + SVM).