Skip to main content
. 2022 Jun 9;9:251. doi: 10.1038/s41597-022-01307-4

Table 5.

Per-pixel confusion matrix of Non-Experts to Expert Consensus.

Expert Consensus
Water Trees Grass Flooded Vegetation Crops Shrub & Scrub Built Area Bare Ground Snow & Ice Cloud Precision/User’s
Non-Experts Water 7814103 36668 21205 58174 193344 18921 17533 68692 3688 3002 94.90%
Trees 237858 16664442 499046 301831 657282 1834363 133325 71652 18987 19774 81.50%
Grass 9921 60697 540597 55648 265055 113899 14984 18240 550 1407 50.00%
Flooded Vegetation 255910 107576 23721 879482 40311 33552 9357 23270 0 0 64.00%
Crops 13392 150401 258153 7920 12567377 223368 92983 88258 0 5141 93.70%
Shrub & Scrub 72746 1748087 1008622 519267 1656386 5143322 245498 1996929 137415 12010 41.00%
Built Area 10716 222831 40154 2028 324127 59196 6224787 35812 866 1510 89.90%
Bare Ground 22661 25522 23480 4482 268845 257830 22460 1036373 49328 4090 60.40%
Snow & Ice 2644 59334 539 687 73 21734 0 15917 1388854 11 93.20%
Cloud 3033 8988 198 264 19887 244 3740 393 3955 161082 79.80%
Recall/Producer’s 92.60% 87.30% 22.40% 48.10% 78.60% 66.70% 92.00% 30.90% 86.60% 80.20% 77.80%

Note that cloud is included as both sets of annotations include this label (n = 58,963,662).