ClassificationEvaluator claims class does not exist, although it is present

[Gigaszi]

When running python vl3d.py --pipeline predict_mixing.json with this file predict_mixing.json, the following plots are created. The CLI also claims that the "Tree" class does not exist. However, Tree is displayed in the Confusion Matrix and the class is also contained in the predicted point cloud.

[Gigaszi]

I can send the data privately

[albertoesmp]

Hi @Gigaszi

First, thank you for using the new VL3D framework. We strongly appreciate your feedback.

Could you let me know which branch you are using? I guess it is the main one, right?

Best,
Alberto

[Gigaszi]

Hi,
yes it was the main branch. I used the mine_transform_and_train_pipeline_pca_from_url.json for the training of the model.

[albertoesmp]

I think I found the problem. The classification labels start at 1, but they should start at 0. After updating the classification labels to start at zero, I got the class distribution plot below.

I hope this solves the issue. Otherwise, please let me know.