Table 3 Median of differences between one-stage models and two-stage models (1s and 2s represent one-stage model and two-stage model respectively)
From: Semantic segmentation and photogrammetry of crowdsourced images to monitor historic facades
IoU (%) | F1 (%) | MPA (%) | Precision (%) | Recall (%) | |
|---|---|---|---|---|---|
2s–1s (M1) | 12.2 (2.0, 28.1) | 10.7 (1.2, 22.9) | 0.3 (− 1.5, 2.4) | 14.7 (1.9, 28.6) | 0.4 (− 4.4, 5.2) |
2s–1s (M2) | − 3.5 (− 10.6, 1.6) | − 1.9 (− 9.8, 1.7) | 0.0 (− 0.8, 1.6) | − 5.0 (− 14.3, 1.1) | 0.8 (− 1.2, 5.3) |
2s–1s (M3) | 4.9 (− 3.0, 9.2) | 3.6 (− 1.7, 7.9) | − 0.6 (− 1.9, 0.9) | 7.9 (− 2.5, 12.1) | − 1.3 (− 4.3, 3.4) |