Trained Helsinki no-bottom model
omarsamiranov opened this issue · comments
Hello,
I would like to run the current approach on a custom dataset. My dataset of buildings contain no bottoms for the buildings. You mentioned in the paper that for the real world Shenzhen dataset you used Helsinki no-bottom model. Can you please provide the trained model?
Also I expect that I will use it the same way as provided in the readme. Just adding the model url and name in the config file, am I correct?
Hi @omarsamiranov, one no-bottom model I have just at hand for a follow-up work has been trained on airborne LiDAR. Otherwise, you can easily train one with the provided Helsinki data. Please always mind the gap between the training data and your custom dataset, and retrain it if possible.
The only difference I reckon at the moment when using a no-bottom model is to set append_bottom
to true
:
Line 23 in f194e87