ABSTRACT
This paper presents a web application to control personnel access to a work area without contact;this makes it ideal to help combat the Covid-19 health emergency. For its implementation, deep learning and computer vision techniques have been used for face detection and recognition. The system consists of four phases, the first one aimed at detecting and aligning the face with deep learning algorithms. The second phase obtains the facial features to recognize different people. The third phase consists of implementing a module that detects face impersonation, and significantly prevents possible attacks on the system by identifying whether the face is real or fake;and the last phase is the design and development of the web interface. This interface performs the communication of the algorithms, the users and the administration. In order to evaluate this proposal, several experiments have been carried out under diverse real conditions. The main results to correctly identify the user show that it has an accuracy of 99 %, in an estimated time of 3 seconds, in the range of 20 cm to 90 cm away, with respect to the camera. In addition, the system is capable of identifying users wearing masks or glasses, in this case with an accuracy of 95% in 4 seconds. © 2022 IEEE.
ABSTRACT
Due to the coronavirus pandemic and the lack of an automatic COVID-19 diagnostic system to relieve congestion in health centers and to support the traceability of this disease, this article exposes the implementation of algorithms for automatic diagnosis of lung diseases such as COVID-19 and Pneumonia from chest X-rays (CXR) through GLCM and HOG features extraction using 6300 patches. Then, selecting the best features and different classifiers such as an Support Vector Machine (SVM) and Artificial Neural Network (ANN) to obtain a system maximum accuracy of 93,73% for SVM. © 2021 IEEE.