Neighbourhood low-pass filters (5 × 5 kernel size) were applied to remove outlier pixels and speckling.
Effects of kernel size on grain sorghum [Sorghum bicolor (L).
Moench] quality were studied in an experiment designed to separate effects of kernel size from seedlot.
Kernel size effects on Rapid Visco Analyzer (RVA) properties were not consistent.
We also present an adaptive kernel size selection method instead of traditional manual selection.
Chemical composition, physical characteristics, milling characteristics, pasting properties, and cooking qualities were determined for each kernel size fraction.
Results: The spatial resolution of SWS estimates was directly related to the reconstruction kernel size in TOF methods, and the precision increased with increasing kernel size.
G1.5 mm indicates Gaussian filter of 1.5-mm kernel size; G2.5 mm, Gaussian filter of 2.5-mm kernel size.
Therefore, we use two convolutional layers in this paper; in terms of the kernel size, the results for a 3 × 1 kernel size are nearly identical to those for a 5 × 1 kernel size.
Numbers written on left side of conv, pool, fc indicate [(#kernel) × kernel size, width × height) / (stride)], [(#kernel size, width × height) / (stride)], [(#output node)] respectively.
Therefore, one would use a light (1.5-mm kernel size) smoothing rather than a heavy (2.5-mm kernel size) one in view of measurement reproducibility.
