Abstract
The recent Covid-19 pandemic and the fact that wearing masks in public is now mandatory in several countries, created challenges in the use of face recognition systems (FRS). In this work, we address the challenge of masked face recognition (MFR) and focus on evaluating the verification performance in FRS when verifying masked vs unmasked faces compared to verifying only unmasked faces. We propose a methodology that combines the traditional triplet loss and the mean squared error (MSE) intending to improve the robustness of an MFR system in the masked-unmasked comparison mode. The results obtained by our proposed method show improvements in a detailed step-wise ablation study. The conducted study showed significant performance gains induced by our proposed training paradigm and modified triplet loss on two evaluation databases.

Abstract

Abstract

Abstract

Abstract
Pollutant emissions from passenger cars give rise to harmful effects on human health and the environment. Predicting traffic flow is a challenging problem, but essential to understand what factors influence car traffic and what measures should be taken to reduce carbon dioxide emissions. In this work, we developed a predictive model to forecast traffic flow in several locations in the city of Porto for 24 h later, i.e., the next day at the same time. We trained a XGBoost Regressor with multi-modal data from 2018 and 2019 obtained from traffic and weather sensors of the city of Porto and the geographic location of several points of interest. The proposed model achieved a mean absolute error, mean square error, Spearman's rank correlation coefficient, and Pearson correlation coefficient equal to 80.59, 65395, 0.9162, and 0.7816, respectively, when tested on the test set. The developed model makes it possible to analyse which areas of the city of Porto will have more traffic the next day and take measures to optimise this increasing flow of cars. One of the ideas present in the literature is to develop intelligent traffic lights that change their timers according to the expected traffic in the area. This system could help decrease the levels of carbon dioxide emitted and therefore decrease its harmful effects on the health of the population and the environment.

Abstract
Photoplethysmography (PPG) is a noninvasive way to monitor various aspects of the circulatory system, and is becoming more and more widespread in biomedical processing. Recently, deep learning methods for analyzing PPG have also become prevalent, achieving state of the art results on heart rate estimation, atrial fibrillation detection, and motion artifact identification. Consequently, a need for interpretable deep learning has arisen within the field of biomedical signal processing. In this paper, we pioneer novel explanatory metrics which leverage domain-expert knowledge to validate a deep learning model. We visualize model attention over a whole testset using saliency methods and compare it to human expert annotations. Congruence, our first metric, measures the proportion of model attention within expert-annotated regions. Our second metric, Annotation Classification, measures how much of the expert annotations our deep learning model pays attention to. Finally, we apply our metrics to compare between a signal based model and an image based model for PPG signal quality classification. Both models are deep convolutional networks based on the ResNet architectures. We show that our signal-based one dimensional model acts in a more explainable manner than our image based model; on average 50.78% of the one dimensional model's attention are within expert annotations, whereas 36.03% of the two dimensional model's attention are within expert annotations. Similarly, when thresholding the one dimensional model attention, one can more accurately predict if each pixel of the PPG is annotated as artifactual by an expert. Through this testcase, we demonstrate how our metrics can provide a quantitative and dataset-wide analysis of how explainable the model is.