Table 1.

FCN training hyperparameters.

Stage	Hyperparameter	Value
Initialisation	Weights	(i) He et al. [55] (scratch)
		(ii) ImageNet (transfer)
	Bias	0
Dropout	Rate p	0.5
Intermediate Non Linearity Unit	ReLu
Epochs	125 – 150
Optimisation (SGD)	Learning Rate	0.001
	Momentum	0.9
	Decay	0.0016
	Mini Batch	20