Yang7879/3D-BoNet

About small instances in ScanNet.

zhangguanghui1 opened this issue · 1 comments

In your evaluation code, you removed small instances. I would like to ask about the Scannet dataset. How did you handle these small instances when uploading benchmark?
Thank you very much.

Hi @zhangguanghui1 , you are suggested to remove the small ins as well. Use the same function to compute the mean size for each category.

def get_mean_insSize_by_sem(dataset_path, train_areas):