Xiaoyang Wu

Hi @arik1089 , I followed the download command provided by docs and just limited the download file to mesh only. Here is the command ```{bash} python download_data.py raw --split Training...

Hi Guys @ZhengdiYu @xiaobaishu0097, I found a solution. (For Vertices in Area_5b)

Hi @ZhengdiYu: Actually, I noticed that "point_cloud.mat" in Area_5a and Area_5b both saved all data of Area_5, but the coords saved in Area_5a's "point_cloud.mat" correspond with S3DIS Dataset, but all...

More specific descriptions of this bug: Assume self.C and self.F are on 'cuda' -> 'min_coord = min(self.C) is on 'cuda' -> tensor_stride = torch.IntTensor() is on 'cpu' -> Boom.

I never encounter the same situation. I submitted the benchmark twice, one for 73.x and another is 75.x. We also discussed the phenomenon, and some senior researchers told me that...

Do you mean issue #9 ? The example config in issue #9 was a config from an older version of our codebase and these are base configs that experiment config...

Hi, could you show me your full script and I will check it? There is no `alignmentAngle.txt` in the S3DIS aligned angle version of the raw dataset.

Hi, this repo is no longer maintain, please considering shift to our new codebase, Pointcept.

Hi, could you show me your custom dataset config (whole exported config in exp folder is better) and OOM output?

The config looks good. Is there any potential that the validation point clouds are significantly larger than the training point clouds? What about `data_dict["coord"].shape`? (It would be helpful if you...