Table 3.

Confusion matrices for four classification model architectures on test images.

(a) ResNet50
		Predicted
		Unfed	Fully fed	Semi-gravid	gravid
Actual	Unfed	71	0	0	0
	Fully fed	0	49	0	3
	Semi-gravid	0	2	63	1
	Gravid	0	0	0	45

(b) MobileNetV2
		Predicted
		Unfed	Fully fed	Semi-gravid	Gravid
Actual	Unfed	71	0	0	0
	Fully fed	0	45	1	6
	Semi-gravid	1	3	58	4
	Gravid	2	1	2	40

(c) EfficientNet-B0
		Predicted
		Unfed	Fully fed	Semi-gravid	Gravid
Actual	Unfed	71	0	0	0
	Fully fed	0	48	0	4
	Semi-gravid	0	2	60	4
	Gravid	2	0	3	40

(d) ConvNeXtTiny
		Predicted
		Unfed	Fully fed	Semi-gravid	Gravid
Actual	Unfed	65	1	0	5
	Fully fed	0	48	2	2
	Semi-gravid	3	1	52	10
	Gravid	1	2	4	38

Correctly predicted instances are in bold.