During the first 15 years of the twenty-first century, Brazil’s economic growth and public policies were in the center of the debate on the growing “new middle class.” This new middle class is defined by people’s household income between the upper 10th percentile and the median (Neri, A Nova Classe Média, 2008). Although there has been a consensus about the increase in consumption and the improvement of living conditions for a significant proportion of the population, there is less agreement about the decline in inequality and the change in class distribution. Previous work was directed at challenging the very idea that Brazil had become a middle-class country during the first decade of this century, basically weighting class distribution against income distribution. In this article, we aim to step into the income distribution debate using six income groups as proportions of the median household per-capita income. Our data source is the National Household Sample Survey (PNAD-IBGE/Brazil) in 2001, 2008, and 2015. We analyze groups’ income distribution and characteristics using multinomial logistic models to take into account the effects of socioeconomic variables. We argue that there is significant stability in groups’ income structure during the period, revealing their resistance to inequalities (similar to the findings in the works of Piketty and Souza). We also indicate that the odds of being included in the upper-income categories are quite unequal, considering socioeconomic variables. Finally, we point out that the gains observed from 2001 to 2008 had faded by 2015 when the odds of being included in the upper-income categories were remarkably similar to those of 2001.
Nosúltimos anos, os avanços em Aprendizado Profundo revolucionaram diversas subareas da Visão Computacional, incluindo o Rastreamento de Objetos Visuais. Um tipo especial de rede neural profunda, a Rede Neural Siamesa, chamou a atenção da comunidade especializada em rastreamento. Ela possui baixo custo computacional e alta eficácia para comparar a similaridade entre objetos. Atualmente, a comunidade científica atingiu resultados notáveis ao aplicar tais redes ao problema de Rastreamento de Objetos Visuais.No entanto, observou-se que limitações dessa rede neural impactam negativamente no rastreamento. Superou-se o problema ao se obter um novo descritor para referência do objeto combinando descritores passados fornecidos pelo rastreador. Em particular, foi proposto a combinação de sinal de descritores em blocos de memórias de longo e de curto prazo, os quais representam a primeira e a mais recente aparência do objeto, respectivamente.Um descritor finalé gerado a partir desses blocos de memória, o qual o rastreador usa como referência. Este trabalho enfatizou-se na obtenção de um método para calcular um banco de filtros otimizado através do uso de um algoritmo genético. O banco de filtroś e utilizado então para gerar a saída da memória de curto prazo. De acordo com experimentos realizados na base de dados OTB, esta proposta apresenta ganhos em comparação com a proposta original da SiamFC. Considerando a métricaárea abaixo da curva, há ganhos de 7.4% e 3.0% para os gráficos de precisão e sucesso, respectivamente, tornando este trabalho comparável a métodos do estato da arte.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.