Xiaoyang Wu

I am sorry about that issue. I never encounter a similar problem. I notice that the validation batch size per GPU is identical to the train batch size per GPU...

Hi, that was quite a huge number for a point cloud after voxelization. Maybe you can further validate whether the validation point cloud voxelized successfully.

Hi, this issue https://github.com/Gofinge/PointTransformerV2/issues/23 might be helpful \.

Hi, here is our code for S3DIS visualization (Yixing's code), just for reference: ```python import open3d as o3d from pathlib import Path import torch import numpy as np from matplotlib...

Hi, maybe you can compute the mIoU of each Scene, then select a low mIoU scene and check visualization. ```python data_root = Path("data") / "s3dis" / "Area_5" scene_paths = sorted(list(data_root.glob("*.pth")))...

Hi, how about shifting to our new codebase Pointcept? As PCR for PTv2 is no longer maintained.

I haven’t encountered a similar issue before. Hope someone else can help you with the issue.

Yes, this dataset exist a problem, download link is available in our new codebase readme.

> Is PointCept？ ok I konw, thank you yes

Sure, here is the download [link](https://connecthkuhk-my.sharepoint.com/:u:/g/personal/wuxy_connect_hku_hk/Ed3wwHgE5RdHqBRuejIvxdkBxtcRtYS4h4cbIFANU9Tbqw?e=gIE72T) (for the released config). According to feedback from issues, the released 3-stage config is not stable. Maybe a 4-stage config as the original paper...