Table 1.

Main training hyperparameters used for SSD and YOLOv3 networks.

Hyperparameter	SSD	YOLOv3
Training epochs	176,000	5000
Batch size	Between 6 and 10 images	Between 4 and 8 images
Optimizer	RMSProp	SGD Burn-In (in DarkNet [])
Learning rate	[0.001, 0.004]	[0.0001, 0.01]
Momentum	0.9	[0.8, 0.9]
Decay	0.9	[0.0003, 0.0005]