2020/10/30 by Roshan Reddy Yedla, Yedla, Roshan Reddy, Shiv Ram Dubey +1
Computer Science · Neuroscience · #Advanced Neural Network Applications #Brain Tumor Detection and Classification #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Neural Networks and Applications
paper · pdf · doi:10.48550/arxiv.2011.06496
openalex publication_date 2020/10/30 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
Convolutional neural networks (CNNs) have shown very promising performance in\nrecent years for different problems, including object recognition, face\nrecognition, medical image analysis, etc. However, generally the trained CNN\nmodels are tested over the test set which is very similar to the trained set.\nThe generalizability and robustness of the CNN models are very important\naspects to make it to work for the unseen data. In this letter, we study the\nperformance of CNN models over the high and low frequency information of the\nimages. We observe that the trained CNN fails to generalize over the high and\nlow frequency images. In order to make the CNN robust against high and low\nfrequency images, we propose the stochastic filtering based data augmentation\nduring training. A satisfactory performance improvement has been observed in\nterms of the high and low frequency generalization and robustness with the\nproposed stochastic filtering based data augmentation approach. The\nexperimentations are performed using ResNet50 model over the CIFAR-10 dataset\nand ResNet101 model over Tiny-ImageNet dataset.\n