Skip to main content
. 2021 Jan 11;11:559172. doi: 10.3389/fpls.2020.559172

TABLE 5.

Comparison of results from ImageNet- and PlantNet-pre-trained backbones in a single-stage detector.

Class Average precision of the ImageNet pre-trained backbone Average precision of PlantNet
Angular leafspot 74.72% 81.69%
Anthracnose fruit rot 87.38% 90.23%
Gray mold 89.68% 90.00%
Leaf blight 87.64% 89.81%
Leaf spot 71.66% 80.48%
Powdery mildew fruit 83.94% 84.01%
Powdery mildew leaf 86.91% 88.59%
mAP 83.13% 86.40%