Figure 1 - uploaded by António J. R. Neves
Content may be subject to copyright.
![Figure 2. Gradient associated with a region of an RGB image [15].](profile/Antonio-Neves/publication/325569674/figure/fig2/AS:779763010506754@1562921397816/Gradient-associated-with-a-region-of-an-RGB-image-15_Q320.jpg)
![Figure 3. Matrix for certain area of a grayscale image [17].](profile/Antonio-Neves/publication/325569674/figure/fig3/AS:779763010506755@1562921397951/Matrix-for-certain-area-of-a-grayscale-image-17_Q320.jpg)
![Figure 4. Application of a Bayer filter to obtain a color image [19].](profile/Antonio-Neves/publication/325569674/figure/fig4/AS:779763010527233@1562921397976/Application-of-a-Bayer-filter-to-obtain-a-color-image-19_Q320.jpg)
![Figure 5. Example of CCD (left) and CMOS (right) sensor [20].](profile/Antonio-Neves/publication/325569674/figure/fig5/AS:779763014713344@1562921398003/Example-of-CCD-left-and-CMOS-right-sensor-20_Q320.jpg)
Contexts in source publication
Context 1
... the joining of many pixels and due to the filtering effect of the human eye, it is possible to create illusions, like gradients and shading. Figure 1 shows an array of pixels relating to a digital image and Figure 2 represents the gradient of a Red, Green and Blue (RGB) image by merging pixels. The most common types of digital images are grayscale and RGB images. ...
Context 2
... body detection, one of two data sets are used: Common Objects in Context (COCO) or MPII Human pose dataset, with people images, annotated with the human skeleton, still being used CMU Panoptic dataset during the development of the algorithm, since it contains about 65 sequences of approximately 5 hours and 30 minutes and 1.5 million 3D skeletons available. This detection is done through the approach described in [31], where a neural network is used to simultaneously predict confidence maps for body part detection ( Figure 10b) and affinity fields for association of parts of the body (see Figure 10c), this process being done in several steps, so that this detection is credible. ...
Context 3
... body detection, one of two data sets are used: Common Objects in Context (COCO) or MPII Human pose dataset, with people images, annotated with the human skeleton, still being used CMU Panoptic dataset during the development of the algorithm, since it contains about 65 sequences of approximately 5 hours and 30 minutes and 1.5 million 3D skeletons available. This detection is done through the approach described in [31], where a neural network is used to simultaneously predict confidence maps for body part detection ( Figure 10b) and affinity fields for association of parts of the body (see Figure 10c), this process being done in several steps, so that this detection is credible. ...
Context 4
... a set of two-part combinations is performed to associate the body parts, where a score is used to define which person belongs to the respective part and to make a correct connection of the parts in each person in the image/frame (Figure 10d). Through this approach, it is possible to detect several people in the image and define their posture. ...
