misatch between gt_labels and seg_preds

[D3xter1922]

I am trying to train on my own dataset, and I am using PointSample during in my test and val pipeline. However, the gt_labels are still the labels of the entire input, and the predictions are still the predictions of the sampled points. This leads to an error in the evaluation step. Any help on this would be appreciated.

Thanks

[filaPro]

You can try to sample point with fixed random seed, and then use the same seed to sample annotation points during metric calculation.

[D3xter1922]

Thank you for your reply. Could you tell me what changes do I need to make in my val_evaluator in my config, and how to specify seed for PointSample_ ? Or the only option is to hardcode it in the metric calculation pipeline.