Xiaoyang Wu

Results 495 comments of Xiaoyang Wu

Oh, yes. Some point clouds might be padded as `drop_last=False.` In the final test script, we maintain a dictionary to prevent duplicate point clouds involved in metric computation. But for...

@SpeedyGonzales949 Thanks for your answer.

Hi, please use PTv3 and PTv2 instead. We didn't optimized PTv1 for outdoor lidar dataset.

Sorry, no preprocessing for this dataset. Free to open a PR if you write one.

@blueeaglex Did you use the official config we released?

Hi, I have never encountered this issue before.

Install the environment with the following environment.yml. ``` name: pointcept-torch2.3.1-cu12.1 channels: - pyg - pytorch - nvidia/label/cuda-12.1.1 - nvidia - bioconda - conda-forge - defaults dependencies: - python=3.10 - pip...

Hi, could you show me your config file? Maybe you forgot this in the first line: https://github.com/Pointcept/Pointcept/blob/main/configs/scannet/semseg-pt-v3m1-0-base.py#L1

Hi, how about bs12 with gpu4? Do you have any other baseline, and align the augmentation with these baseline?

Hi, use PPT joint training on semantic tasks with nuScenes, SemanticKITTI, and Waymo. And then tune the model on your task (choose a prompt from any dataset that is close...