Pre-training Point-BERT and dVAE on other datasets

[eliahuhorwitz]

Hey,
In the runner and runner_BERT_pretrain scripts there is a check for dataset name and if it's not ShapeNet (or ModelNet for runner_BERT_pretrain), a NotImplementedError is thrown.
What do I need to change/add/remove in order to run the pertaining and dVAE on other datasets (ScanObjectNN or ScanNet for example)?

Thanks,
Eliahu

[yuxumin]

Hi,
Thanks for your interest in our work.

1. Create your dataset class under datasets directory. And remember to put the class into registry.
2. Initialize the custom dataset class by importing it in datasets/__init__.py
3. Change the dataset config in certrain method config.
4. Modify the code in tools/runner.py. ( check the dataset name and preprocess the data for model)

Then you can run the code on your own dataset.

reference: issue 11

[yuxumin]

Close it since no response. Feel free to re-open it if problems still exist