Including background class for training

[rozeappletree]

Checklist

• [X] I have searched for similar issues.
• [X] I have tested with the latest development wheel.
• [X] I have checked the release documentation and the latest documentation (for master branch).

My Question

Hi,

as in current config file we are ignoring background label.

https://github.com/isl-org/Open3D-ML/blob/8ddb67206e4fef55b39eea691ff00d49cef18be5/ml3d/configs/randlanet_toronto3d.yml#L6-L7

1. Is it the reason why we are adding all class labels by 1 during inference in inference script below? If not, what is the reason? I could not understand it.

https://github.com/isl-org/Open3D-ML/blob/8ddb67206e4fef55b39eea691ff00d49cef18be5/examples/vis_pred.py#L41

2. And what does Fill "unlabeled" value because predictions have no 0 values. mean here below? https://github.com/isl-org/Open3D-ML/blob/8ddb67206e4fef55b39eea691ff00d49cef18be5/examples/vis_pred.py#L47-L48 It seems, we are just assigning the zero-th index of prediction labels 0. What does it even mean?

3. I want to train a model with only one class. What do you think is the best way to proceed for the task

   • Ignore background class and train on single class? (or)
   • Include background class and train it as binary predictor - that will give two probabilities is it an asset & is it not an asset

[QasimMuhammad]

@INF800 hello, did you find the answer? Will appreciate if you share