Hello,
I am working on TIDL_J7_01_00_00_00. I am attempting to import a neural network which accepts RGB planar input and takes normalized input tensors.
In the training framework, my colleague applied the Z-Score normalization process (range [-1, 1]) using the following steps:
(1) image = image(in Float32) / 255.0
(2) image = image - mean
(3) image = image / standard_deviation
The specified mean values for each channel are in the range [0,1]. But am not sure how to specify the scaling values in this case.
Please advise.
Thank you.
