DA-ActNN-YOLOV5: Hybrid YOLO v5 Model with Data Augmentation and Activation of Compression Mechanism for Potato Disease Identification

Table 2

Accuracy of the DA-ActNN-YOLOV5m model using different data augmentation methods in the PBD-IM dataset.

No.	Data augmentation methods	Precision (%)			mAP (%)
		Earlyblight	Healthy	Lateblight
1	Rotation_transform	96.72	97.56	96.43	97.56
	Width_transform_range
	Height_transform_range
2	Rotation_transform	98.87	98.44	97.85	98.39
	Width_transform_range
	Height_transform_range
	Shear_transform
	Zoom_transform
3	Rotation_transform	99.12	100.00	98.94	99.75
	Width_transform_range
	Height_transform_range
	Shear_transform
	Zoom_transform
	Horizontal_flip
	Brightness_transform
	Channel_transform_range
	Fill_nearest
4	Rotation_transform	99.98	100.00	99.85	99.81
	Width_transform_range
	Height_transform_range
	Shear_transform
	Zoom_transform
	Horizontal_flip
	Brightness_transform
	Channel_transform_range
	Fill_nearest
	Random_rain
	Random_fog
	Random_shadow
	Shadow_roi