clustering result is always be like 'count': 0

[MisuZeng]

May I ask if you are considering releasing the extracted point clouds for your synthetic and real-world data. From Table I and Fig.8. At present, I have reached the clustering step, but the clustering count is always 0. Through debugging, I found that clustering has failed (isinstance (X, int) and X == -1). And the semantic_comlormap.ply obtained from volume sampling cannot be opened with a 3D viewer, suspecting that there may be a problem with volume sampling. I would like to download your extracted point clouds to investigate clustering issues, or can you give me any suggestions based on my problem description?

Thank you very much! Wishing you a smooth work!

[meyerls]

Oh yes, I have originally planned to published the point cloud used for evaluation. I will upload them and notify you if it is finished.

[MisuZeng]

Ok. Thank you so much!

[park-dea]

hello, has the point cloud been released?

[Cola-1]

@meyerls hello, has the point cloud been released?

[yimoW]

I can open the semantic_comlormap.ply through meshlab, but the clustering result is still 0. I've met the same issue as listed

[EyGy]

Have a look at my answer for #21
did you try lowering the values for remove_outliers_nb_points and remove_outliers_radius?
Perhaps the outlier removal steps are not correctly configured to match your data...

[park-dea]

@EyGy I conducted experiments on both real data and synthetic data (FruitNerf dataset) by gradually lowering the parameters for remove_outliers_nb_points and remove_outliers_radius, but it still results in a count of 0.

[EyGy]

@park-dea
Well if you get 0 using the synthetic FruitNerf Dataset and the respective parameters out of the "config_synthetic.py" that is strange, unexpected behavior. Without access to your code this will be difficult to debug and probably is caused by your setup?

Can you double check, whether you use the correct parameters to the respective synthetic dataset? And are your paths correct?

If yes, start debugging with custom print messages in the "clustering_base.py". Somewhere between loading the dataset and the DBSCAN clustering steps there will be an error probably caused by using wrong parameters or paths - e.g. use print messages to check upon the lenght of the loaded pcd during the preprocessing and clustering.