Transfer learning based cascaded deep learning network and mask recognition for COVID-19.

The COVID-19 is still spreading today, and it has caused great harm to human beings. The system at the entrance of public places such as shopping malls and stations should check whether pedestrians are wearing masks. However, pedestrians often pass the system inspection by wearing cotton masks, scarves, etc. Therefore, the detection system not only needs to check whether pedestrians are wearing masks, but also needs to detect the type of masks. Based on the lightweight network architecture MobilenetV3, this paper proposes a cascaded deep learning network based on transfer learning, and then designs a mask recognition system based on the cascaded deep learning network. By modifying the activation function of the MobilenetV3 output layer and the structure of the model, two MobilenetV3 networks suitable for cascading are obtained. By introducing transfer learning into the training process of two modified MobilenetV3 networks and a multi-task convolutional neural network, the ImagNet underlying parameters of the network models are obtained in advance, which reduces the computational load of the models. The cascaded deep learning network consists of a multi-task convolutional neural network cascaded with these two modified MobilenetV3 networks. A multi-task convolutional neural network is used to detect faces in images, and two modified MobilenetV3 networks are used as the backbone network to extract the features of masks. After comparing with the classification results of the modified MobilenetV3 neural network before cascading, the classification accuracy of the cascading learning network is improved by 7%, and the excellent performance of the cascading network can be seen.

Development of Face Mask Detection using SSDLite MobilenetV3 Small on Raspberry Pi 4

This study aimed to develop a mask detection tool with SSDLite MobilenetV3 Small based on Raspberry Pi 4. SSDLite MobilenetV3 Small is a single-stage object detection. The single-stage object detection method is faster than the two-stage detection method. However, it has the disadvantage as the level of accuracy is not as good as the two-stage detection method. In the experiments, we used some methods to compare with SSDLite MobilenetV3, such as: SSDLite MobilenetV3 Large, SSDLite MobilenetV2, SSD MobilenetV2, SSDLite Mobileedets, and SSDMNV2 models. The result is that SSDLite MobilenetV3 is more powerful than other systems for detecting face masks. While the model with the best detection is the SSDLite MobilenetV2 model, the system with the SSDLite MobilenetV3 Small model still detects the use of masks, with a score of 70% accuracy from model accuracy testing in deployment. The limitation is the system with SSDLite MobilenetV3 Small can't detect incorrect masks. © 2022 IEEE.

An Efficient Face Mask Wearing Detection Algorithm Based on Improved YOLOv3

In recent years, the Corona Virus Disease 2019 (Covid-19) epidemic has raged around the world, with more than 500 million people diagnosed. Relevant medical research and analysis results on Covid-19 indicate that wearing masks is an effective method to prevent and restrain virus transmission. Mask detection stations have been set up in hospitals, railway stations, schools, where there is large crowd flow, but results are not as good as expected. In order to ameliorate pandemic preventing and control measures, a mask wearing detection algorithm YOLOv3-M3 was designed and proposed in this paper. The algorithm can effectively detect people without mask, while consequently reminding them. Firstly, we substituted the feature extraction network of YOLOv3 with MobileNetv3, a lightweight convolutional neural network. Secondly, we utilized K-Means++ to substitute the original ground truth clustering algorithm to improve prediction precision. In addition, the bounding box regression loss function was revised as CIoU loss function. This loss function solves the issues of overlapping between the ground truth and the anchor box, which has increased the training speed. After experiments, the precision of YOLOv3 algorithm on mAP 0.5 and mAP 0.75 is 93.5% and 71.9%, respectively. Elevating 3.1% and 2.6%, respectively, higher than that of YOLOv3 algorithm, and it was superior to SSD, SSD Lite, YOLOv3-Tiny and other one-stage object detection algorithms. The detection speed can reach 13.6 frame/s, which has met the requirements of pandemic prevention and control in most places and can be deployed on terminal devices for object detection. [ FROM AUTHOR]

The COVID-19 Images Classification by MobileNetV3 and Enhanced Sine Cosine Metaheuristics

The research proposed in this paper shows the application of the MobileNetV3 and improved sine cosine algorithm utilized for solving problems of optimization, thus including COVID-19 image classification. The algorithm that is the subject of optimization is a more recent solution in the field of population-based algorithms. The algorithm initializes random solutions upon each new iteration, and these solutions are candidates for the optimal solution. Solutions should generally tend toward the best solution or the opposite from it, depending in the problem to be solved. Such operations are performed by utilizing the sine and cosine functions in a mathematical adaption for the problem of optimization. The proposed improved version of the algorithm has been used to address the COVID-19 image classification problem. The problem of this case is a very challenging one and the study verifies and demonstrates the proposed improvements in terms of algorithm’s performance. The obtained findings from the proposed improved algorithm suggest superior performances in contrast to other methods included in the research. The proposed algorithm outperforms any other metaheuristics, in particular in terms of the feature numbers and classification accuracy. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

A Deep Learning Approach for Face Mask Detection

The whole world is passing through a very difficult time since the outbreak of Covid-19. Wave after wave of this pandemic hitting people very hard across the globe. We have lost around 3.8 million lives so far to this pandemic. Moreover, the impact of this pandemic and the pandemic-induced lockdown on the lives and livelihoods of the people in the developing world is very significant. Till now there is no one-shot remedy available to stop this pandemic. However, spread can be controlled by social distancing, frequent hand sanitization, and using a face mask in public places. So, in this paper, we proposed a model to detect face mask of people in public places. The proposed model uses OpenCv module to pre-process the input images, it then uses a deep learning classifier MobileNetV3 for face mask detection. The accuracy of the proposed model is almost 97%. The proposed model is very light and can be installed on any mobile or embedded system. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

Transfer Learning-based Mobile-focused Automated COVID-19 Detection from Chest X-ray

COVID-19 Pandemic is still a global issue that threatens global health. To combat the pandemic, testing activities has been the first line of defense. However, increasing number of infections resulted in insufficient number of laboratory kits to perform the test. One potential testing method is using transfer learning for automated detection of COVID-19 from chest x-ray image. We create a model used pretrained model of MobileNetV3Large as a feature extractor, and a custom classification layer. We train the model on dataset consisting of chest x-ray image from 10,192 healthy cases, 3,616 COVID-19 cases, 1,345 Viral Pneumonia cases, and 6,012 Lung Opacity cases. The model achieved macro-average accuracy performance of 89.08%, F1 score of 88.10%, Precision of 91.95%, Sensitivity of 85.51%, and Specificity of 95.26%. Comparison with previous models trained on smaller dataset showed that achieved performance is lower and indicates previous research's model won't be able to maintain its performance when evaluated on larger sets of data. © 2021 IEEE.