Skip to main content
. 2019 Dec 31;9:20387. doi: 10.1038/s41598-019-56309-x

Table 2.

The mean square error and mean truncated relative error of the various CNN models.

Regular CNN trained using the loss function Equ. 7 Regular CNN trained using the loss function Equ. 10 Porosity-informed CNN Regular CNN with preprocessed image as input
Mean square error 8.64 × 10−4 7.01 × 10−4 7.64 × 10−4 6.92 × 10−4
Mean truncated relative error 68.8% 41.7% 59.2% 29.7%
R2 0.9903 0.9921 0.9914 0.9912