Fit prediction for filtering facepiece respirator using 3D face shape.

Ensuring proper respirator fit for individuals remains a persistent challenge in occupational environments, yet there is limited knowledge about how respirators interact with the face to "'fit." Previous studies have attempted to understand the association between face dimensions and respirator fit using traditional head/face anthropometry not specifically tailored for respirators. The purpose of this study was to assess and compare the ability of filtering facepiece respirator (FFR)-specific face anthropometry with traditional head/face anthropometry in exploring the relationship between facial dimensions and the fit of FFR. The study utilized 3D face scans and quantitative fit factor scores from 56 participants to investigate the relationship between face anthropometry and FFR fit. Both FFR-specific and traditional anthropometric measurements were obtained through 3D anthropometric software. Intra-correlation of anthropometry was analyzed to evaluate the efficiency and effectiveness of FFR-specific and traditional anthropometry respectively. Principal component analysis (PCA) was conducted to test the usefulness of the PCA method for investigating various facial features. Logistic regression was used to develop fit association models by estimating the relationship between each face measurement set and the binary outcome of the fit test result. The prediction accuracy of the developed regression models was tested. FFR-specific face anthropometry consists of a set of measurements that can inform the detailed facial shape associated with the FFRs more effectively than traditional head/face anthropometry. While PCA may have been effective in reducing the variable dimensions for the relatively large parts of the human body such as upper and lower bodies in previous literature, PCA results of FFR-specific and traditional anthropometry were inconsistent and insufficient to describe face dimensions with complex anatomy in a small-detailed area, suggesting that facial shape should be understood through a variety of approaches including statistical methods. Logistic regression analysis results confirmed that the association models of FFR-specific face anthropometry were significant with higher prediction accuracy and had a better model's goodness of fit than those of traditional head/face anthropometry in 3 conditions inputting all measurements, all PC scores, or top 5 measurements from PCA. The findings showed that the FFR fit association model enables an understanding of the detailed association between face and respirator fit and allows for the development of a system to predict respirator fit success or failure based on facial dimensions. Future research would include testing the validity of the model and FFR-specific measurement set on different respirator types, expanding the population set, and developing an integrated approach using automated and machine learning technologies to inform FFR selection for occupation workers and the general population.

3D Face modeling is still one of the biggest challenges in computer graphics. In this paper we present a novel framework that acquires the 3D shape, texture, pose and illumination of a face from a single photograph. Additionally, we show how we can recreate a face under varying illumination conditions. Or, essentially relight it. Using a custom-built face scanning system, we have collected 3D face scans and light reflection images of a large and diverse group of human subjects. We derive a morphable face model for 3D face shapes and accompanying textures by transforming the data into a linear vector sub-space. The acquired images of faces under variable illumination are then used to derive a bilinear illumination model that spans 3D face shape and illumination variations. Using both models we, in turn, propose a novel fitting framework that estimates the parameters of the morphable model given a single photograph. Our framework can deal with complex face reflectance and lighting environments in an efficient and robust manner. In the results section of our paper, we compare our methods to existing ones and demonstrate its efficacy in reconstructing 3D face models when provided with a single photograph. We also provide several examples of facial relighting (on 2D images) by performing adequate decomposition of the estimated illumination using our framework.

The importance of fit testing in decontamination of N95 respirators: A cautionary note

In this work, different methods are presented to create 3D face models from databases of 3D face scans. The challenge in this endeavour is to balance the limited training data with the high demands of various applications. The 3D scans stem from various persons showing different expressions, with varying number of points per 3D scan and different numbers of scans per person. This data of posed facial expressions revealed substructures, which are utilised to improve the proposed model. In the process of creating and using the models, for each specifc application objective quality criteria are carefully designed tailored to the task to quantify the quality. In total four face models built from three databases are compared based on: 3D face synthesis, 3D approximation, person and expression transfer, and 3D reconstruction from 2D. Contents Abbreviations and Nomenclature XII 1 Introduction 1 1.1 The Difficulty of Quality Assessment . . . . . . . . . . . . . . 2 1.2 Face Models ....

SARS-CoV-2 is a respiratory virus. Transmission occurs by droplets, contact and aerosols. In medical settings, filtering facepiece (FFP) respirators are recommended for use by personnel exposed to aerosol-generating procedures. During the COVID-19 pandemic, the demand for FFP respirators exceeded their supply worldwide and low-quality products appeared on the market, potentially putting healthcare workers at risk. To raise awareness about variations in quality of imported FFP respirators in Switzerland during the COVID-19 pandemic, to draw attention to the current directives regulating the market launch of FFP respirators in Switzerland, to provide practical support in identifying suspicious products or documents and, finally, to offer strategies aimed at reducing the distribution of low-quality FFP respirators in the future. Three Swiss laboratories, Spiez Laboratory and Unisanté in partnership with TOXpro SA individually set up testing procedures to evaluate aerosol penetration and fit testing of FFP respirators imported into Switzerland during COVID-19 pandemic. Additionally, Spiez Laboratory visually inspected the products, examined the certification documents and crosschecked the product information with international databases. Between 31 March and 15 June 2020, 151 FFP respirators were analysed. The initial assessment performed before testing allowed a reduction of up to 35% in the number of FFP respirators sent to Spiez Laboratory for evaluation, for which product information found to be faulty. After filtration efficiency evaluation and fit testing, 52% and 60% of all products tested by Spiez Laboratory and Unisanté-TOXpro SA, respectively, did not meet the minimum performance requirements established independently by the three Swiss laboratories. The demand for FFP respirators exceeded the supply capacity from established suppliers of the Swiss market. New production and import channels emerged, as did the number of poor-quality FFP respirators. FFP respirators remaining in stocks should be checked for conformity before being used, or eliminated and replaced if quality does not meet standards.

This thesis presents an approach to the modeling of facial aging using and extending the Morphable Model technique. For modeling the face variation across individuals, facial expressions, and physical attributes, we collected 3D face scans of 298 persons. The 3D face scans where acquired with a structured light 3D scanner, which we improved in collaboration with the manufacturer to achieve superior geometry and texture quality. Moreover, we developed an efficient way to measure fine skin structure and reflection properties with the scanner. The collected face scans have been used to build the Basel Face Model, a new publicly available Morphable Model. Using the 3D scans we learn the correlation between physical attributes such as weight, height, and especially age and faces. With the learned correlation, we present a novel way to simultaneously manipulate different attributes and demonstrate the capability to model changes caused by aging. Using the attributes of the face model in conjunction with a skull model developed in the same research group, we present a method to reconstruct faces from skull shapes which considers physical attributes, as the body weight, age etc. The most important aspect of facial aging that can not be simulated with the Morphable Model is the appearance of facial wrinkles. In this work we present a novel approach to synthesize age wrinkles based on statistics. Our wrinkle synthesis consists of two main parts: The learning of a generative model of wrinkle constellations, and the modeling of their visual appearance. For learning the constellations we use kernel density estimation of manually labeled wrinkles to estimate the wrinkle occurrence probability. To learn the visual appearance of wrinkles we use the fine scale skin structure captured with our improved scanning method. Our results show that the combination of the attribute fitting based aging and the wrinkle synthesis, facilitate a simulation of visually convincing progressive aging. The method is without restrictions applicable to any face that can be represented by the Morphable Model.

Deep networks trained on millions of facial images are believed to be closely approaching human-level performance in face recognition. However, open world face recognition still remains a challenge. Although, 3D face recognition has an inherent edge over its 2D counterpart, it has not benefited from the recent developments in deep learning due to the unavailability of large training as well as large test datasets. Recognition accuracies have already saturated on existing 3D face datasets due to their small gallery sizes. Unlike 2D photographs, 3D facial scans cannot be sourced from the web causing a bottleneck in the development of deep 3D face recognition networks and datasets. In this backdrop, we propose a method for generating a large corpus of labeled 3D face identities and their multiple instances for training and a protocol for merging the most challenging existing 3D datasets for testing. We also propose the first deep CNN model designed specifically for 3D face recognition and trained on 3.1 Million 3D facial scans of 100K identities. Our test dataset comprises 1,853 identities with a single 3D scan in the gallery and another 31K scans as probes, which is several orders of magnitude larger than existing ones. Without fine tuning on this dataset, our network already outperforms state of the art face recognition by over 10%. We fine tune our network on the gallery set to perform end-to-end large scale 3D face recognition which further improves accuracy. Finally, we show the efficacy of our method for the open world face recognition problem.

Three-dimensional object construction has seen a great deal of activity in the past decade, as has been pointed out in recent surveys, with the advancement of technology and easy availability of the depth sensors the 3D scanning technology has taken off. A wide range of commercial sensors such as Intel RealSense, Microsoft Kinect are being widely used for real time 3D capturing. Efficient 3D face scanning is one of the important areas of 3D scanning. 3D printer compatible texture supported scanning has a wide range of commercial applications. Such methods are also being used for 3D avatars and characterizations for games. Even though several commercial grade applications are available, most of the techniques suffer from background and light variations. Therefore an efficient face scanning technique is of extreme importance. In this paper we propose a 3D face scanning method based on Intel RealSense technology that combines 3D face detection, background segmentation and 3D mesh mapping to produce realistic 3D face model along with texture export. Further the models then can be manipulated using any 3D editing software along with texture and wireframe manipulation.

In the process of designing custom prostheses, there are various traditional or digital techniques to measure the limbs of the patients. The most widely used method is traditional anthropometric measurements. On the other hand, the development of three-dimensional imaging methods in the digital environment such as computed tomography (CT), magnetic resonance imaging (MRI), and 3-dimensional (3D) scanning have made these applications more practical and precise. However, the deviations and measurement differences that will affect the measurement results among them have not been clearly revealed yet. Therefore, measurements have been collected with MRI, 3D scanning and traditional anthropometric measurement methods from a person that has an amputation on his hand and the results have been compared. The results were statistically analyzed with one-way ANOVA and Bland-Altman plots, and no significant difference was found between the measurement methods (p<0.98). Although the 3D scanning is close to the MRI results, the lowest deviation was obtained in the MRI method. As a result, 3D scanning is expected to be used more widely in the future in the custom design phase in terms of cost-effectiveness, time-saving, measurement reliability, and repeatability.

BackgroundMillions of people rely on N95 filtering facepiece respirators to reduce the risk of airborne particles and prevent them from respiratory infections. However, there are no respirator fit testing and training regulations in China. Meanwhile, no study has been conducted to investigate the fit of various respirators. The objective of this study was to investigate whether people obtained adequate fit when wearing N95 filtering facepiece respirators (FFRs) used widely in China.MethodsFifty adult participants selected using the Chinese respirator fit test panel donned 10 common models of N95 FFRs. Fit factors (FF) and inward leakage were measured using the TSI PortaCount Plus. Each subject was tested with three replications for each model. A subject was considered to pass the fit test when at least two of the three FFs were greater than 100. Two models were conducted fit tests before and after training to assess the role of training.ResultsThe geometric mean FFs for each model and trained subjects ranged from <10 to 74.0. The fifth percentile FFs for only two individual respirator models were greater than 10 which is the expected level of performance for FFRs. The passing rates for these two models of FFRs were 44.7% and 20.0%. The passing rates were less than 10.0% for the other eight models. There were 27 (54%) participants who passed none of the 10 FFRs. The geometric mean FFs for both models when the subjects received training (49.7 and 74.0) were significantly larger than those when the same group of subjects did not receive any training (29.0 and 30.9) (P<0.05).ConclusionsFFRs used widely in China should be improved according to Chinese facial dimensions. Respirator users could benefit from respirator training and fit testing before using respirators.

This study compared the simulated workplace protection factors (SWPFs) between NIOSH-approved N95 respirators and P100 respirators, including two models of filtering facepiece respirator (FFR) and two models of elastomeric half-mask respirator (EHR), against sodium chloride particles (NaCl) in a range of 10 to 400 nm.Twenty-five human test subjects performed modified OSHA fit test exercises in a controlled laboratory environment with the N95 respirators (two FFR models and two EHR models) and the P100 respirators (two FFRs and two EHRs). Two Scanning Mobility Particle Sizers (SMPS) were used to measure aerosol concentrations (in the 10–400 nm size range) inside (Cin) and outside (Cout) of the respirator, simultaneously. SWPF was calculated as the ratio of Cout to Cin. The SWPF values obtained from the N95 respirators were then compared to those of the P100 respirators.SWPFs were found to be significantly different (P<0.05) between N95 and P100 class respirators. The 10th, 25th, 50th, 75th and 90th percentiles of the SWPFs for the N95 respirators were much lower than those for the P100 models. The N95 respirators had 5th percentiles of the SWPFs > 10. In contrast, the P100 class was able to generate 5th percentiles SWPFs > 100. No significant difference was found in the SWPFs when tested against nano-size (10 to 100 nm) and large-size (100 to 400 nm) particles.Overall, the findings suggest that the two FFRs and two EHRs with P100 class filters provide better performance than those with N95 filters against particles from 10 to 400 nm, supporting current OSHA and NIOSH recommendations.

As a classic statistical model of 3D facial shape and albedo, 3D Morphable Model (3DMM) is widely used in facial analysis, e.g., model fitting, image synthesis. Conventional 3DMM is learned from a set of 3D face scans with associated well-controlled 2D face images, and represented by two sets of PCA basis functions. Due to the type and amount of training data, as well as, the linear bases, the representation power of 3DMM can be limited. To address these problems, this paper proposes an innovative framework to learn a nonlinear 3DMM model from a large set of in-the-wild face images, without collecting 3D face scans. Specifically, given a face image as input, a network encoder estimates the projection, lighting, shape and albedo parameters. Two decoders serve as the nonlinear 3DMM to map from the shape and albedo parameters to the 3D shape and albedo, respectively. With the projection parameter, lighting, 3D shape, and albedo, a novel analytically-differentiable rendering layer is designed to reconstruct the original input face. The entire network is end-to-end trainable with only weak supervision. We demonstrate the superior representation power of our nonlinear 3DMM over its linear counterpart, and its contribution to face alignment, 3D reconstruction, and face editing. Source code and additional results can be found at our project page: http://cvlab.cse.msu.edu/project-nonlinear-3dmm.html.

As a classic statistical model of 3D facial shape and texture, 3D Morphable Model (3DMM) is widely used in facial analysis, e.g., model fitting, image synthesis. Conventional 3DMM is learned from a set of well-controlled 2D face images with associated 3D face scans, and represented by two sets of PCA basis functions. Due to the type and amount of training data, as well as the linear bases, the representation power of 3DMM can be limited. To address these problems, this paper proposes an innovative framework to learn a nonlinear 3DMM model from a large set of unconstrained face images, without collecting 3D face scans. Specifically, given a face image as input, a network encoder estimates the projection, shape and texture parameters. Two decoders serve as the nonlinear 3DMM to map from the shape and texture parameters to the 3D shape and texture, respectively. With the projection parameter, 3D shape, and texture, a novel analytically-differentiable rendering layer is designed to reconstruct the original input face. The entire network is end-to-end trainable with only weak supervision. We demonstrate the superior representation power of our nonlinear 3DMM over its linear counterpart, and its contribution to face alignment and 3D reconstruction.

Classification of skeletal discrepancies by machine learning based on three-dimensional facial scans.

Accurate registration of cone-beam computed tomography scans to 3-dimensional facial photographs