Table 4 Comparison of the experimental results before and after improvement
From: GrainNet: efficient detection and counting of wheat grains based on an improved YOLOv7 modeling
Models | Variety | Accuracy | Efficiency | FPS | Inference time | |||||
|---|---|---|---|---|---|---|---|---|---|---|
P (%) | R (%) | F1 | mAP@ 0.5 (%) | FLOPs (G) | Params (M) | Model size (M) | ||||
YOLOv7 | G | 96.89 | 87.11 | 0.918 | 91.03 | 103.20 | 36.49 | 73.06 | 32.36 | 21.64 |
Y | 95.87 | 91.89 | 0.940 | 93.30 | ||||||
B | 95.36 | 86.46 | 0.908 | 90.74 | ||||||
ALL | 96.04 | 88.49 | 0.921 | 91.69 | ||||||
GrainNet | G | 97.52 | 92.67 | 0.950 | 93.58 | 87.40 | 30.54 | 50.66 | 29.10 | 22.36 |
Y | 97.95 | 92.56 | 0.952 | 93.53 | ||||||
B | 96.50 | 90.91 | 0.936 | 92.35 | ||||||
ALL | 97.32 | 92.05 | 0.946 | 93.15 | ||||||