CNN + OBIA
I'm following the deep learning tutorial on ecog TV. Now the tutorial recommends to stick to 32 bit float or 8bit. I'm sure my data isn't rising to the occasion because I'm getting this warning.
I'm using planetscope 3M data and I stitched the imagery on ArcGIS then imported into Ecog. Ground truth points too, are collected through ArcGIS online. Do I prep my img and thematic layers differently?
Was this article helpful?
2 comments
Hi Lynn,
For the CNN algorithms to work in eCognition your input data needs to be in 8bit or 32bit float. You simply can convert your input layers to one of these formats using the "layer arithmetics" algorithm. You can create a copy of your input layers and define the bit depth in the parameter section of the algorithm.
Please also see here for more information about the "Layer Arithmetics" algorithm:
Cheers,
Matthias
Matthias, thank you so much. I did that and it worked.
I had to do the same process when applying the CNN model too and it looks like it's working (still processing). What a relief. Thank you!