Throwaway Shadows Using Parallel Encoders Generative Adversarial Network

Face photographs taken on a bright sunny day or in floodlight contain unnecessary shadows of objects on the face. Most previous works deal with removing shadow from scene images and struggle with doing so for facial images. Faces have a complex semantic structure, due to which shadow removal is challenging. The aim of this research is to remove the shadow of an object in facial images. We propose a novel generative adversarial network (GAN) based image-to-image translation approach for shadow removal in face images. The first stage of our model automatically produces a binary segmentation mask for the shadow region. Then, the second stage, which is a GAN-based network, removes the object shadow and synthesizes the effected region. The generator network of our GAN has two parallel encoders—one is standard convolution path and the other is a partial convolution. We find that this combination in the generator results not only in learning an incorporated semantic structure but also in disentangling visual discrepancies problems under the shadow area. In addition to GAN loss, we exploit low level L1, structural level SSIM and perceptual loss from a pre-trained loss network for better texture and perceptual quality, respectively. Since there is no paired dataset for the shadow removal problem, we created a synthetic shadow dataset for training our network in a supervised manner. The proposed approach effectively removes shadows from real and synthetic test samples, while retaining complex facial semantics. Experimental evaluations consistently show the advantages of the proposed method over several representative state-of-the-art approaches.

People who inject drugs in metropolitan Chicago: A meta-analysis of data from 1997-2017 to inform interventions and computational modeling toward hepatitis C microelimination

Progress toward hepatitis C virus (HCV) elimination in the United States is not on track to meet targets set by the World Health Organization, as the opioid crisis continues to drive both injection drug use and increasing HCV incidence. A pragmatic approach to achieving this is using a microelimination approach of focusing on high-risk populations such as people who inject drugs (PWID). Computational models are useful in understanding the complex interplay of individual, social, and structural level factors that might alter HCV incidence, prevalence, transmission, and treatment uptake to achieve HCV microelimination. However, these models need to be informed with realistic sociodemographic, risk behavior and network estimates on PWID. We conducted a meta-analysis of research studies spanning 20 years of research and interventions with PWID in metropolitan Chicago to produce parameters for a synthetic population for realistic computational models (e.g., agent-based models). We then fit an exponential random graph model (ERGM) using the network estimates from the meta-analysis in order to develop the network component of the synthetic population.

Health and wellbeing needs and priorities in mining host communities in South Africa: a mixed-methods approach for identifying key SDG3 targets

Background The global mining industry has an opportunity to mobilize resources to advance progress against the Sustainable Development Goals (SDGs). In 2018, the Anglo-American Group outlined aspirations for mining host communities to meet the SDG3 health targets. To progress from aspiration to action we designed and implemented a mixed-methods approach to attain a deeper understanding of the health and wellbeing priorities within the local context of host communities of fifteen mines in South Africa. Methods To identify local needs and priorities relating to SDG3 targets in host communities, stakeholder workshops and key informant interviews were conducted between June and August 2019. A baseline assessment of health data, related to each of the SDG3 targets and indicators and to each host community location, was also conducted. Findings emerging from the qualitative and quantitative baseline assessments were compared to identify the extent to which health issues aligned and health and wellbeing priority areas for action. Results A total of 407 people participated in the workshops, and 85 key informants were interviewed. Quantitative data were available at sub-national level for seven of the nine SDG3 targets and eleven of the 21 indicators. Key priority areas for action identified through alignment of the qualitative and quantitative data were maternal mortality (SDG3.1), HIV (SDG3.3.1), tuberculosis (SDG3.3.2), substance abuse (SDG3.5), and road traffic accidents (SDG3.6) We found consistency in the individual, interpersonal, community, societal, and structural factors underlying these priority areas. At a structural level, poor access to quality healthcare was raised at every workshop as a key factor underlying the achievement of all SDG3 targets. Of the five priority areas identified, HIV, TB and substance abuse were found to overlap in the study communities in terms of risk, burden, and underlying factors. Conclusions We demonstrate a mixed method approach for identifying local health needs and prioritised SDG3 targets in mining host communities. Consistency in reporting suggests the need for effective, efficient and feasible interventions to address five priority areas. Given the prominent economic role of the mining sector in South Africa, it can play a critical role in implementing programmatic activities that further progress towards achieving the SDG3 targets.

How can a Climate-Neutral Building Look Like?

The climate crisis is urging us to act fast. Buildings are a key leverage point to reduce greenhouse gas (GHG) emissions, but the embodied emissions related with their construction remain often the hidden challenge of any ambitious policy. Considering that a complete material substitution is not possible, we explore in this paper a material GHG compensation where fast-growing bio-based insulation materials are used to compensate building elements that necessarily release GHG. Looking for analogies with other human activities, different material diets as well as different building typologies are modelled to assess the consequences in term of bio-based insulation requirement to reach climate-neutrality. The material diets are defined according to the gradual use of herbaceous materials, from the insulation up to the structural level: omnivorous, vegetarian and vegan. Our results show the relationship in terms of volume between the climate intensive materials and the climate-negative ones needed to neutralize the overall building GHG emissions. Moreover, they suggest how climate-neutral building can look like and that it is possible to have climate-neutral buildings with wall thickness within the range of current construction practices.

Radiation-doped SiC*/Si heterostructure formation and defects evolution

The authors consider heterostructures of silicon carbide obtained during endotaxy on silicon substrates. The question is raised in connection with the description of the endotaxy process itself at the structural level. Authors focus on the technological aspects of the formation of a stable β-SiC/Si heterostructure by endotaxy in relation to the evolution of point defects of various nature and their probable association models with the participation of a radionuclide impurity at the micro-alloying level: 1) the growth of the SiC*/Si thin layer with C-14 atoms in the doping process; 2) physical properties of defects formation; 3) some interface between properties and efficiency.

Invisible or Powerful? Ageing in a Mediatised Society

In order to live your rights and achieve your goals, you need to be informed, have a voice and be listened to, and have the opportunity to engage in society, regardless of age. Freedom of expression and freedom of information are core human rights values that connect the concept of capability with the role of the media in society. The media can be a tool for enhancing a person’s capabilities, but it can also be seen as hindering a good life if technology and its applications are perceived as awkward and/or difficult to access.At a structural level, a wider discussion of media responsibility has the potential to contribute to enhancing people’s capabilities in later life. One critical issue is who should be held accountable and responsible for media content that lacks diverse stories about older people and their voices, possibly reinforcing ageism. Furthermore, where the responsibility lies for ensuring that older people have the technological means to act as digital citizens is somewhat unclear. In an increasingly mediatised environment, we might see a stronger relationship between media literacy, health and ageing, which in turn could emphasise the importance of the role of the media in enhancing capabilities.

Analisis Kesalahan dalam Menyelesaikan Soal Bentuk Aljabar Ditinjau dari Taksonomi SOLO

The aim of this study is to describe the types of errors and their causes made by the seven grade students of SMPN 7 PUJUT in the academic year 2020/2021 in solving algebraic form questions based on the SOLO taxonomy level. This research is a qualitative descriptive study and the instruments used are test questions and interview guides. The subjects of this study were 32 grade seven students of SMPN 7 PUJUT who were selected using purposive sampling technique. After selecting one class, several students were selected according to the SOLO taxonomy level. The results showed that all students made the types of errors according to the basic objects of mathematics, namely fact errors, concept errors, operating errors, and principle errors. The results also showed that the questions given to 32 class seven students of SMPN 7 PUJUT obtained a percentage of the structural level of 31.25%, the unistructural level of 37.5%, the multistructural level of 21.9%, the relational level of 9.38%, and the extended abstract level is 0%. The causes of students making mistakes in solving questions, namely: the cause of factual errors is that students do not understand the meaning of the questions which include what is known and asked and in determining the final result; the cause of misconceptions is that students do not understand the formula used; the cause of the operation error is that the students are not careful in calculating; and the cause of the principle error is that students do not understand the steps to solve it.

Comparative analysis of Lysine and Arginine biosynthesis pathway in Deinococcus genomes

The members of the Deinococcaceae family have the ability to survive extreme environmental conditions. Deinococcus species have a complex cell envelope composed of L-ornithine containing peptidoglycan. Anabolism of L-ornithine is intrinsically linked to L-lysine and L-arginine biosynthetic pathways. To understand these two pathways, we analyzed the L-lysine and L-arginine pathways using 23 Deinococcus genomes, including D. indicus. We used BLAST-P based ortholog identification using D. radiodurans genes as the query. We identified some BLAST-P hits that shared the same functional annotation. We analyzed three (class I aminotransferase, acetyl-lysine deacetylase, and acetyl glutamate/acetyl aminoadipate kinase) from L-lysine biosynthesis pathway and three (bifunctional ornithine acetyltransferase or N-acetyl glutamate synthase protein, nitric oxide synthase-like protein, and Acetyl-lysine deacetylase) from L-arginine biosynthesis pathway. Two proteins showed certain structural variations. Specifically, [LysW]-lysine hydrolase protein sequence and structure level changes indicated changes in oligomeric conformation, which could likely be a result of divergent evolution. And, bifunctional ornithine acetyltransferase or N-acetyl glutamate synthase had its active site pocket positions shifted at the structural level and we hypothesize that it may not perform at the optimal level. Thus, we were able to compare and contrast different Deinococcus species indicating some genes occurring because of divergent evolution.

Optimized convolutional neural network architectures for efficient on-device vision-based object detection

Convolutional neural networks have pushed forward image analysis research and computer vision over the last decade, constituting a state-of-the-art approach in object detection today. The design of increasingly deeper and wider architectures has made it possible to achieve unprecedented levels of detection accuracy, albeit at the cost of both a dramatic computational burden and a large memory footprint. In such a context, cloud systems have become a mainstream technological solution due to their tremendous scalability, providing researchers and practitioners with virtually unlimited resources. However, these resources are typically made available as remote services, requiring communication over the network to be accessed, thus compromising the speed of response, availability, and security of the implemented solution. In view of these limitations, the on-device paradigm has emerged as a recent yet widely explored alternative, pursuing more compact and efficient networks to ultimately enable the execution of the derived models directly on resource-constrained client devices. This study provides an up-to-date review of the more relevant scientific research carried out in this vein, circumscribed to the object detection problem. In particular, the paper contributes to the field with a comprehensive architectural overview of both the existing lightweight object detection frameworks targeted to mobile and embedded devices, and the underlying convolutional neural networks that make up their internal structure. More specifically, it addresses the main structural-level strategies used for conceiving the various components of a detection pipeline (i.e., backbone, neck, and head), as well as the most salient techniques proposed for adapting such structures and the resulting architectures to more austere deployment environments. Finally, the study concludes with a discussion of the specific challenges and next steps to be taken to move toward a more convenient accuracy–speed trade-off.

Students' Thinking Level in Solving Mathematics Problems Based on SOLO Taxonomy as Viewed from the Mathematics Anxiety

The SOLO (Structure of Observed Learning Outcome) taxonomy is an educational taxonomy suitable for organizing various types of learning. The SOLO taxonomy categorizes students' thinking into five levels: pre-structural, uni-structural, multi-structural, relational, and extended abstract. The purpose of this study was to describe the level of students' thinking in solving mathematics problems based on the SOLO taxonomy with high, medium, and low levels of mathematics anxiety. This type of research is descriptive qualitative research. This research was conducted in one of the junior high schools in Tulungagung City, East Java, Indonesia. The instruments used were a mathematics anxiety questionnaire, test based on the SOLO taxonomy, and interview guidelines. The data analysis used the Miles and Huberman model, which consists of three stages, namely data reduction, data presentation, and conclusion drawing or verification. The results showed that subjects with high mathematics anxiety had a uni-structural level of thinking. Second, subjects with moderate mathematics anxiety had a multi-structural level of thinking. Third, subjects with low mathematics anxiety have an extended abstract thinking level.