Table 1.

YieldNet architecture.

Type/stride	Padding	Filter size	Number of Filters
Conv/s2	Valid	$7\times 7$	48
Conv/s2	Valid	$5\times 5$	64
Conv/s2	Same	$5\times 5$	96
Conv/s1	Same	$3\times 3$	128
Conv/s1	Same	$3\times 3$	128
Corn head
Conv/s1	Same	$3\times 3$	148
Conv/s1	Same	$3\times 3$	148
FC-100
FC-50
Soybean head
Conv/s1	Same	$3\times 3$	148
Conv/s1	Same	$3\times 3$	148
FC-100
FC-50

The first five convolutional layers work as a backbone feature extractor and share weights for both corn and soybean yield predictions.