Table 2.

Deep learning hyperparameters.

Hyperparameter	Value
Optimizer [55]	Stochastic gradient descent (SGD) with learning rate $ϵ=0.00001$, momentum $\mu =0.9$, L2 weight decay $\alpha =0.0001$
Epochs	25
Training examples	5000
Batch size	8
Image size	640 pixel × 360 pixel