Table 11 Performance evaluation of PFLO across maize growth stages and occlusion levels
From: PFLO: a high-throughput pose estimation model for field maize based on YOLO architecture
Growth stage | Occlusion level | Model | Precision (%) | Recall (%) | mAP50 (%) | mAP50 - 95 (%) | OKS | PCK@0.2 | ||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
Box | Pose | Box | Pose | Box | Pose | Box | Pose | |||||
V3-V5 | Minimal (0–20%) | Baseline | 88.2 | 73.4 | 86.5 | 70.8 | 91.5 | 70.1 | 61.0 | 38.1 | 0.5582 | 0.8073 |
PFLO | 88.9 | 75.9 | 88.3 | 72.6 | 92.5 | 72.4 | 65.9 | 40.8 | 0.5703 | 0.8180 | ||
V6-V8 | Moderate (20–40%) | Baseline | 87.5 | 77.4 | 87.8 | 74.3 | 93.0 | 77.1 | 67.1 | 44.9 | 0.5549 | 0.8402 |
PFLO | 88.7 | 79.9 | 89.1 | 77.8 | 93.5 | 79.6 | 70.1 | 48.1 | 0.5765 | 0.8563 | ||
V9-R1 | Significant (40–60%) | Baseline | 81.4 | 66.9 | 82.5 | 63.2 | 87.6 | 65.0 | 61.3 | 36.4 | 0.4804 | 0.7699 |
PFLO | 83.7 | 71.5 | 85.3 | 66.9 | 88.9 | 68.3 | 64.2 | 39.5 | 0.5040 | 0.7914 | ||