Closed SC-shendazt closed 1 year ago
Hi, it should be possible. The only thing that is necessary is to check how the clustered instances look like (output from this function). You can then adjust the hyperparameters from the plane fitting and from DBSCAN to extract good clusters from the S3DIS data.
Hi, it should be possible. The only thing that is necessary is to check how the clustered instances look like (output from this function). You can then adjust the hyperparameters from the plane fitting and from DBSCAN to extract good clusters from the S3DIS data. Hopefully you can update how to change the loading of the custom dataset to fit the Segcontrast pipeline, such as S3DIS
On our new paper we have a new method for pre-training and in there we have some tips on how to implement a new dataloader for the model: https://github.com/PRBonn/TARL
Can segcontrast be implemented on S3DIS without changing a lot of code?