@mlscale attribute
![:electric_plug: :electric_plug:](https://github.githubassets.com/images/icons/emoji/unicode/1f50c.png)
dp.kinect3
dp.oak
-
signature
mlscale FLOAT [FLOAT]...
-
examples
@mlscale 0.429 0.424 0.425
Scale input values after subtracting mean: (pixel_value - mean) * scale
.
Some models are trained so that inference later needs standardized input.
Use @mlmean
and @mlscale
to standardize real-time input for these models.
The given @mlscale
floats are multiplied to each channel in the same order.
3-channel input with @mlswapch
applies @mlscale
to channels 2,1,0 and swaps data
channels 0 and 2 before sending to the model as input.