Table 6 Evaluation of PFLO metrics by class and dataset split
From: PFLO: a high-throughput pose estimation model for field maize based on YOLO architecture
Split | Class | mAP50 (%) | mAP50 - 95 (%) | OKS | PCK@0.2 | ||
|---|---|---|---|---|---|---|---|
Box | Pose | Box | Pose | ||||
Train | Stalk | 97.1 | 90.5 | 77.0 | 65.5 | 0.692 | 0.923 |
Leaf | 94.0 | 86.3 | 77.7 | 61.5 | 0.657 | 0.896 | |
Validation | Stalk | 94.2 | 74.6 | 65.6 | 44.1 | 0.514 | 0.837 |
Leaf | 91.0 | 76.9 | 71.0 | 49.8 | 0.551 | 0.827 | |
Test | Stalk | 92.2 | 69.4 | 63.1 | 39.4 | 0.500 | 0.803 |
Leaf | 90.8 | 75.0 | 68.9 | 46.0 | 0.537 | 0.815 | |