Input to predict.py is not in the same dimension as the output

[OmarAhmadin]

hi,

So I have done training with a good results.I did not use any C++ preprocessing for the dataset but I preprocessed the dataset using some tools my self. Now, for prediction, I want to feed let's say 1 scene which has let's assume: NPOINTS and I get the prediction with the same size of NPOINTS in the same order so I can just save the input data with the predicted labels and get the original file but with the new labels.

Any idea how to do that?

predict.py always generates NPOINTS =8192 all the time!

[rjose97]

Hello

Were you able to solve it? Did you use predict.py with scannet? There are certain data in the scannet dataset class that are not supported.

Thanks in advance. Greetings.