Removing Zero Variance Units of Deep Models for COVID-19 Detection

Deep Learning has been used for several applications including the analysis of medical images. Some transfer learning works show that an improvement in performance is obtained if a pre-trained model on ImageNet is transferred to a new task. Taking into account this, we propose a method that uses a pre-trained model on ImageNet to fine-tune it for Covid-19 detection. After the fine-tuning process, the units that produce a variance equal to zero are removed from the model. Finally, we test the features of the penultimate layer in different classifiers removing those that are less important according to the f-test. The results produce models with fewer units than the transferred model. Also, we study the attention of the neural network for classification. Noise and metadata printed in medical images can bias the performance of the neural network and it obtains poor performance when the model is tested on new data. We study the bias of medical images when raw and masked images are used for training deep models using a transfer learning strategy. Additionally, we test the performance on novel data in both models: raw and masked data. Author

Face Mask Detection for Preventing the Spread of Covid-19 using Knowledge Distillation

The coronavirus pandemic (COVID-19) has unfolded hastily throughout the entire world. This pandemic disease can spread through droplets and can be airborne. Hence, the use of face masks in public places is crucial to stop its spread. The present study aims to develop a system that can identify masked or non-masked faces;whether it is a normal mask, transparent mask, or a face alike mask. The face mask detection system is developed with the help of Convolutional Neural Networks (CNN). The model compression technique of Knowledge Distillation has been used to make the machine lesser computation and memory intensive so that it is simple to install the model on a few embedded gadgets and cell computing platforms. Using the model compression technique and GPU systems will help boom the calculation velocity of the model and drop the storage space required for calculations. The experimental outcomes show that the developed detector is capable to classify diverse types of masks. Also, it can classify video images in real-time. Using the Knowledge Distillation on the baseline model can improve the testing accuracy from 88.79% to 90.13%. The proposed unique system can be implemented to assist in the prevention of COVID-19 spread and detect various mask types. © 2021 IEEE.