Lloyd Hughes

I'm experiencing the same issue

I have solved this using the following steps: 1) `git clone https://github.com/myriadrf/gr-limesdr.git` into any directory 2) `source /Applications/GNURadio.app/Contents/MacOS/usr/bin/grenv.sh` 3) `cd gr-limesdr` 4) `mkdir build & cd build` 5) `cmake ..`...

1) See Figure 5 in the paper – there is an upsampling operation which needs to be added to the output of the network. This is not required during training...

I could share the pretrained models, but I haven't done so yet. It should be noted that as the model was pretrained on TerraSAR-X and PRISM, it would be naive...

That would probably be the easier approach, just check if the entire chunk is masked and if so then cull it.

I'd suggest `is_nodata` or something along those lines to abstract it further away from any specific value (for instance, in my application, these values are often `nan`). To avoid any...

@hansmohrmann fair point. I think taking that into account we could design this to optimize both for the case where all values are `zero` as this continues to hold mathematically,...

Thanks for the detailed response. Basically the spheres are "irrelevant" in terms of shading and material properties. They are just used to the ray tracer has something to intersect. I...

Any updates on this? Especially cropping to reduce the FOV (i.e. crop to 4:3 aspect ratio)

Sure I don't mind waiting, if you can document the process briefly perhaps then I can do it for other weights for other network sizes if I need to, and...