Table 1 Model performance evaluation with Raw dataset under 416 × 416 pixels’ resolution.

From: Tomato detection based on modified YOLOv3 framework

Methods

Activation

FDL

Precision (%)

Recall (%)

F1 (%)

AP (%)

YOLOv3

Leaky

 × 3

97.4

96.2

96.8

97.8

YOLOv4

Mish + SPP

 × 3

97.4

100.0

98.7

99.5

YOLO-Tomato-A

Leaky

 × 3

96.1

99.3

97.7

98.2

YOLO-Tomato-B

Mish

 × 1

96.2

99.4

97.8

99.3

YOLO-Tomato-C

Mish + SPP

 × 2

97.0

99.3

98.1

99.5