What does this implement/fix? Explain your changes.
This PR adds quantization of extracted features. It will also add explicit dtypes, and refactor some of the feature classes to be a bit less interdependent.
Enabling quantization will be automatic if the extractor object is instantiated with dtype='uint8' or dtype='uint16'. Otherwise, it will cast to the given float type ('float32' by default, but it could be float16 or float64 as well.)
Reference Issue
103
104
What does this implement/fix? Explain your changes.
This PR adds quantization of extracted features. It will also add explicit dtypes, and refactor some of the feature classes to be a bit less interdependent.
Enabling quantization will be automatic if the extractor object is instantiated with
dtype='uint8'ordtype='uint16'. Otherwise, it will cast to the given float type ('float32'by default, but it could befloat16orfloat64as well.)