Depthwise Convolution and space to Batch conversion

Is depthwise convolution in TensorRT implemented making use of space to Batch conversion?
I am aware that Dilated convolutions make use of SpaceToBatch conversion in tensorflow.
Can anyone provide an insight into how depthwise layer is implemented in TensorRT.
I am trying to implement MobileNetv1 in TRT 5.0 and it is working fine. Wanted to know more about the internal working of depthwise convolution.