I’m trying to understand how to normalize some data I am reading from a LIDAR sensor. It is not standard image data. I’ve had lots of luck converting the data to images by scaling it to 0-255.
But I’m not longer constrained to use 255 bit range because I’m not converting it to images, instead I am feeding the raw data directly into a databunch and using that with a learner.
The data itself has an average of 750 and a standard deviation of 30.
Should I just convert it to floats from -1 to 1 with (data - data.mean()) / data.std()?
Should I convert it to floats 0 to 1 with something similar.
Should I just leave it as ints around 750?
Thanks