Normalizing sensor data

Hi @all ,

I just tried training a model for Punch/Flex gesture and was not able to understand why are we adding 4 and dividing by 8 similarly for 2000 and 4000 for gx values:

tensor += [
(df['aX'][index] + 4) / 8,
(df['aY'][index] + 4) / 8,
(df['aZ'][index] + 4) / 8,
(df['gX'][index] + 2000) / 4000,
(df['gY'][index] + 2000) / 4000,
(df['gZ'][index] + 2000) / 4000
]

If I capture my own custom data will these value change if yes then how we can feed in the right values for normalising the input tensor

Thanks

Start from the tensor-building expression shown in the issue and trace where the sensor ranges and normalization constants are defined. Compare those assumptions with the data-collection path for custom inputs, then document how the constants should be selected and what normalized values are expected when the explanation is complete.

Written by the indexing model from the issue text.