From: Active learning with point supervision for cost-effective panicle detection in cereal crops
Sampling method | \(\hat{y}^1\) | \(\hat{y}^2\) | \(\hat{y}^3\) | Queried sample |
---|---|---|---|---|
Least confidence | 1–0.5 = 0.5 | 1–0.5 = 0.5 | 1–0.5 = 0.5 | \(\hat{y}_1\) or \(\hat{y}_2\) or \(\hat{y}_3\) |
Margin | 0.5–0.2 = 0.3 | 0.5–0.2 = 0.3 | 0.5–0.2 = 0.3 | \(\hat{y}_1\) or \(\hat{y}_2\) or \(\hat{y}_3\) |
Entropy | 1.86 | 1.66 | 1.96 | \(\hat{y}_3\) |