The output layer consists of two neurons corresponding to the two classes E (empty space between pillars) and P (pillar tip) with respect to which each pixel and its neighborhood were classified.
Table 1 dimensions of PhC elementary cell Object Radius [μm] Heighy [μm] Material Pillar h p /5 h p Silicon Buffer layer h p /2 h p /5 Silicon dioxide Substrate h p /2 h p /2 Silicon Air volume 20 · h p 20 · h p Air.
The differences between groups in both studies failed to reach statistical significance using parametric testing, although statistical significance was achieved using the Wilcoxon rank-sum test in the PILLAR study (p < 0.05).
A training set of 14 images was created, based on sample bSi-1 focusing on seven positions of pillars (species P) and seven positions between pillars (species E) (Fig. 1a).
Pillar spacing d p = 300 nm, experimental results from Ref. 13 (a); d p = 2 nm, experimental results from Ref. 38 (b); d p = 5 nm, experimental results from Ref. 38 (c); d p = 250 nm, experimental results from Ref. 37 (d).
P: pillar tip; E: empty space between pillars.
As previously explained, in this scaling analysis all dimensions are related to the pillar height h p (see Table 1).
An especial current collector of Ti pillar arrays (Ti Ps) is prepared by a facile hydrothermal etching method.
Also, Fig. 1c, d depict the plan view and the side view of cylindrical pillars, in which r p, h p, and d p are the radii of the pillars, the height of the pillars, and the spacing between the pillars, respectively.
