Abstract

In the review of this article, we are evaluating multi-label classification using the deep learning models, comparing four models in this research, such as CNNs, InceptionV3, VGG16, and ResNet50. So, we are developing this model’s architecture with fewer convolutional layers. Originally, chest X-ray dataset got from Kaggle, in the dataset I have 11 million data points, but we are giving only 7936 datapoints and images (due to the resources of local machine), in the data preprocessing, we  are splitting the dataset into three sub-set, test, validation and training randomly at ratio of 60%, 20% and 20%. We got the best model is InceptionV3, having 12% accuracy on positive datapoints. Using the data augmentation technique with lots of functions like rotation, shear, zoom, width & height shift range, and flipping sample-wise normalization as well.