The FaceSDK allows to speed up the processing by using batches.
Batching is supported on Alignment, Builder and Age&Gender algorithms.
Batch size is the number of images processed "in package". It's indicated when the algorithms are initialized.

Be aware: if the actual images package size on the input is less than indicated batch size, actual processing speed will be the same as the speed of processing of indicated batch size.
if images package size is bigger than indicated batch size, actual processing speed will increase proportionaly.

The maximum batch size depends on the algorithm, its version and model of used GPU device (usually max batch size is 256, 512 or 1024).
It's recommended to use batch size multiple of 16: 16, 32, 64, 128, 256, 512, 1024.