Skip to main content
. 2018 Apr 6;22(7):1193–1202. doi: 10.1017/S1368980018000708

Table 1.

Results from the FCN-8s deep learning model

Pixel accuracy (%) Mean accuracy (%) Mean IU (%) Frequency-weighted IU (%)
Training 93·43 81·51 72·74 89·09
Validation 90·41 65·12 55·26 84·86
Testing 92·18 70·58 61·85 87·57
All 93·33 80·78 71·99 88·95

IU, Intersection over Union.