Currently we convert from uint16 of raw data type to float32 to do all the processing.
Some methods can work with 16-bit data before this conversion is actually needed. It is worth considering if such a strict conversion really necessary early on. Memory estimators will need to take that into account as currently we set the data type strictly to float32 for them.
Currently we convert from uint16 of raw data type to float32 to do all the processing.
Some methods can work with 16-bit data before this conversion is actually needed. It is worth considering if such a strict conversion really necessary early on. Memory estimators will need to take that into account as currently we set the data type strictly to float32 for them.