SCIENTIFIC MACHINE LEARNING WITH CONVOLUTIONAL NEURAL NETWORKS TO PREVENT CONTAGION IN PANDEMIC TIMES

This work implemented the faster_rcnn_resnet_101_coco neural architecture in Python with TensorFlow using Scientific Machine Learning to monitor face mask usage for contagion prevention. Convolutional neural networks were used in a supervised model trained on images captured from students aged 18–35. Training was conducted in environments with and without GPUs, evaluating execution time and prediction accuracy. The predictive model was assessed using the confusion matrix, error, precision, recall, and [Formula: see text]1-score. It was trained on 7200 proprietary images. Based on functional tests, the system was validated.