Table 2 of
Rapp, Mol Vis 2023; 29:197-205. Table 2 of
Rapp, Mol Vis 2023; 29:197-205.
Table 2. The performance by the U-Net to detect real intervention.
| Strength of intervention | Data set where bias was induced by removing on sprout in “group 2” | Data set where bias was induced by removing 2 pixels per sprout in “group 2” | ||||||
|---|---|---|---|---|---|---|---|---|
| Ground truth segmentation | U-Net segmentation | Ground truth segmentation | U-Net segmentation | |||||
| absolute ∆ [pixels] | relative ∆ | absolute ∆ [pixels] | relative ∆ | absolute ∆ [pixels] | relative ∆ | absolute ∆ [pixels] | relative ∆ | |
| strong intervention | 199.4 | 49% | 153.5 | 43.3% | 195,5 | 48.8% | 125.1 | 41.3% |
| medium intervention | 83.8 | 24.8% | 64.1 | 21.4% | 66.2 | 20% | 51.1 | 19.6% |
| weak intervention | 52.8 | 16.1% | 36.6 | 12.9% | 44.4 | 14.3% | 35.8 | 14.7% |