Skip to main content
. 2019 Aug 25;2019:9209727. doi: 10.34133/2019/9209727

Table 3.

Agreement, calculated as (true positives + true negatives)/(number of images), between different experts and between experts and the neural network for classifying as infected or not infected for 4000 subimages from 6 different Erysiphe necator inoculated leaf disk samples (given in Table 4).

% Agreement Expert 1 Expert 2 Expert 3 GPMNet
Expert 1 100 94.8 92.9 91.7
Expert 2 94.8 100 92.3 91.0
Expert 3 92.9 92.3 100 89.3
GPMNet 91.7 91.0 89.3 100