I would like to use one of ESRIs pre-trained deep learning models. It specifies that a 3-band RGB layer should be used as input. However, my RGB layer that I plan to use as input contains a fourth layer, an 'alpha' layer. Do I need to remove this layer in any way before using it as input for the deep learning model? Or can I just use the layer as it is, and trust that the model will work fine with it?
Solved! Go to Solution.
The imagery you use should match the imagery the model was trained on.
You can use a raster function to make a version of the imagery without the fourth band.
This tutorial shows the process:
The imagery you use should match the imagery the model was trained on.
You can use a raster function to make a version of the imagery without the fourth band.
This tutorial shows the process: