Bernhard Glueck

Just a quick question: Has there been any progress on this ? I guess there are two parts, 1. Redirecting properties to engine properties via aspect weaving 2. Creating engine...

I haved used a different loss function ( either RSME or my own, very simple absolute height diff ) I get some results, but they are not very good after...

MxNet ist totally fine ....

That sounds like a plan :-) However i think i will have to with modifying the network architecture directly in java... since we plan to train on quite a few...

I think a common case would be to take a basic Mask-RCNN model from the model zoo ( e.g MxNET ) and use transfer learning to fit it to a...

No apologies neccessary. Thank you for your detailed explanation