Table 7 The comparison of expert data with answers from participants for 4, 10 and 20 classes of percentage built-up, by full agreement, percentage of classifications 1 class higher and 1 class lower than the experts, and the percentage remaining.
From: A crowdsourced global data set for validating built-up surface layers
Agreement (%) | One class higher than experts (%) | One class lower than experts (%) | Remaining (%) | |||||
|---|---|---|---|---|---|---|---|---|
All data | Majority | All data | Majority | All data | Majority | All data | Majority | |
4 classes (25%) | 76.4 | 84.5 | 7.4 | 6.1 | 10.8 | 9.0 | 5.3 | 0.3 |
10 classes (10%) | 56.3 | 68.1 | 13.9 | 10.9 | 17.4 | 17.3 | 12.4 | 3.7 |
20 classes (5%) | 39.4 | 48.9 | 15.8 | 14.1 | 17.9 | 22.8 | 26.9 | 14.2 |
