Antonio Carlos Sobieranski's research while affiliated with Federal University of Santa Catarina and other places

... Choosing the right method can improve the accuracy of emotion recognition [87]. The emotion recognition method of different sensors is described in Figure 10. ...

... Though, to an extent digital holographic microscopy has mended this gap [8][9][10][11][12]; however, this imaging technique is expert and resource intensive with heavy computational load, which necessitates development of an accurate and sensitive numerical reconstruction algorithm for meaningful data processing and results [13]. Moreover, the coexistence of twin images is a continuous challenge with these imaging systems [14][15][16]. ...

... Since LeCun's publication, intensive research has been done in the area, including comparing CNN with regular segmentation algorithms. For instance, recently, Pereira Jr. et al. [9] presented a study in which CNNs show good results even in raw images, i.e., with no pre-processing or normalization. ...

... Regarding accuracy, the concept of source trustworthiness [31] plays large role, and data quality can usually only be improved by collecting additional data: In [32], the authors predict the same task labels with features collected from different sensors as data sources and find that even though the features reflect the same concepts, differences in data quality can have a drastic impact on classification performance. Similarly, [33] finds that different camera sensors severely impact the quality of water turbidity estimation, and [34] observe qualitative differences in depth estimations based on different sensors. A technique for improving accuracy in the absence of trustworthy sources is pointed out by [35], where many cheap labels from noisy labelers are combined. ...

... Because of the improvements in image processing techniques and presentation of low-cost camera devices, many ML models for weed detection and classification have been developed. Table 2 below shows different studies [19][20][21][22][23][24][25][26] which used ML techniques to detect and classify weeds, and presents the extracted features/segmentation method and best result for each study. Table 2. Weed-detection-related work using ML techniques. ...

... To make the process faster, more economical, and reliable, researchers have investigated automated processes for pavement condition evaluation, usually based on computer vision, machine learning, and, more recently, deep learning [6][7][8][9]. In recent years, researchers have reviewed different data acquisition technologies, including 1D-sensors, 2D-sensors, and 3D sensors, to automate pavement conditions [10][11][12][13][14][15]. Commercial solutions for automatic pavement assessment with certain limitations exist; the topic is also a focus of academic research. ...

... Segmentation involves identifying the sets of pixels or voxels that form the tissue of interest [16]. Several reviews have been published reporting medical image segmentation methods, along with the strengths and weaknesses and discussing the challenges and outcomes [14,[17][18][19][20][21][22][23][24][25][26][27][28]. A literature review by Sahiner et al. noted that establishing clinical significance is as important as establishing statistical significance for the research. ...

... Essa tese deu continuidade aos trabalhos desenvolvidos em (RATEKE, 2015;RATEKE et al., 2014;JUSTEN et al., 2016), no contexto do projeto de navegação veicular autônoma 1 do Laboratório de Processamento Digital de Imagens e Computação Gráfica (LAPiX) 2 associado ao Instituto Nacional para Convergência Digital (INCoD) 3 , que contam com outros trabalhos, complementares, no mesmo contexto. ...

... Providing an alternative to the method's complexity O(N 3 ) that makes it unsuitable for real-time execution, the global refinement was achieved offline for the position of all frames. [Linhares et al., 2016] presented another approach using intensity-based non-rigid fine adjustment. It is an offline procedure that consists of minimizing the pairwise SSD metric in an evenly spaced overlapping set of images. ...

... The Mahalanobis distance, in both its linear and higherorder polynomial variations, has been shown to produce better results than linear color-metric approaches such as RGB or CIELab, when employed as a customized color-metric in various segmentation algorithms [22,23,24,25,26,27,28]. ...