| Popis: |
Deep learning, also called universal learning approach is a type of machine learning used to perform classification tasks directly from images, text, or sound. The paper focus on performance analysis of four pre-trained deep learning network for the purpose of classification of brain tumor into four grades. The pre-defined convolutional neural networks (CNN) used are AlexNet, GoogLeNet, InceptionV3 and ResNet-50. These networks are pre-trained on ImageNet dataset. The dataset used here is the MRI images of glioma brain tumor. The skull stripping and data augmentation are the preprocessing steps performed on the glioma MRI images. The network architectures are compared based on various metrics including, accuracy, precision, recall, F1-score and time duration for the training and validation. As per the experiment, out of four networks AlexNet and Resnet50 gives better result in terms of time duration and accuracy respectively. AlexNet gives an accuracy of 92.98% with less time of about 19min, whereas ResNet50 gives better accuracy of 96.05% with a time duration of about 100min. |
