Table 3 Comparison of the detection accuracy between the original YOLOv5 and the proposed method on the wheat seedling dataset with different annotations
From: A method for small-sized wheat seedlings detection: from annotation mode to model construction
Annotation | Training Datasets | Testing Datasets | Method | AP |
|---|---|---|---|---|
Local annotation | Small size | All-size | YOLOv5 Proposed | 74.5% 90.1% |
Medium size | All-size | YOLOv5 Proposed | 63.7% 86.3% | |
Large size | All-size | YOLOv5 Proposed | 30.2% 76.5% | |
Muti-size | All-size | YOLOv5 Proposed | 66.5% 84.6% | |
Global annotation | All-size | All-size | YOLOv5 | 60.2% |