Prediction probability sample extracted from the ten tiles in Table 4. Left: prediction probability of the four models used in this work from Thuruthipilly et al. (2022). The black line shows the threshold probability; every object above 0.8 is considered an ML candidate. Right: total number of ML candidates compared to the TPs with grades 1 and 2 (TEGLIE candidates) for every model. Bar element counts are shown above each bar.