Figure 1 | Scientific Reports

Figure 1

From: Deep learning for irregularly and regularly missing data reconstruction

Figure 1

Network architecture at an example of \(\frac{{N}_{h}}{{2}^{4}}\times \frac{{N}_{w}}{{2}^{4}}\) pixels in the lowest resolution. Nh, Nw, and Nc are the height, width, and the number of channels of the input data, respectively. Blocks show the calculated feature hierarchy. Each green box denotes multiple feature maps, and the number of feature maps (i.e., Fi,i∈[1, 5]) is marked on the right of the box. The height-width-size of a feature map is given around the box. The boxes with the same height have the same number of feature maps. The boxes with the same width indicate the same height-width-size of feature maps. The arrows and the right curly brace denote different operations. Numbers in [[â‹…]] are labelled according to Table 1, which are in line with those shown in Fig. 2.

Back to article page