search

ZhaochongAn / COSeg

[CVPR 2024] This repo contains the code for our paper: Rethinking Few-shot 3D Point Cloud Semantic Segmentation
62 stars 8 forks source link

About Visualization #9

Closed AmingWu closed 2 months ago

AmingWu commented 3 months ago

Dear Authors,

When I run your new code, there exists an error:

微信图片_20240626190853

Could you give me some advice? Thanks.

ZhaochongAn commented 3 months ago

Hi, it seems that the target class you specified might not be included in the test split. Could you share the exact commands you used to run the code?

AmingWu commented 3 months ago

This is the command:

python3 main_fs.py --config config/s3dis_COSeg_fs.yaml test True weight "/data1/zyh/COSeg/trained_model/model/model_last.pth" cvfold 0 train_gpu [0] vis_save_path ./vis forvis 1 data_root /data1/zyh/COSeg/datasets/S3DIS/blocks_bs1_s1/data target_class table

ZhaochongAn commented 3 months ago

Your command shows the issue I mentioned. You are using cvfold 0 as the test split but have specified target_class as table, which is used for cvfold 1, not 0. Therefore, the table index is not defined in the test split (cvfold 0).

AmingWu commented 3 months ago

Could you tell me the cvfold 0 owns concret categories?

ZhaochongAn commented 3 months ago

The concrete category divisions can be found in the dataset class definition: https://github.com/ZhaochongAn/COSeg/blob/75f39091b49ed73b2f06bcf68555986d5d75fa30/util/s3dis_fs.py#L53-L61.

AmingWu commented 3 months ago

Thanks. Since my device does not create window to display the results, could you tell me how to save the results to .ply?

AmingWu commented 3 months ago

Besides, the saved results contain pred.npy, query.npy, querylb.npy, sup.npy, and suplb.npy. Could you tell me their meaning? Which one is used for visualization?

ZhaochongAn commented 3 months ago

The visualization code will visualize all these saved results to show the whole few-shot episode. They correspond to query predictions, query data, query labels, support data, and support labels, respectively.

AmingWu commented 3 months ago

Thanks. If I can directly save these files to .ply, it will be convenient to perform visualization for me. Could you give me some advice? Thanks

AmingWu commented 2 months ago

捕获 你好，这是可视化结果，如何才能获得论文上那种标注特定颜色的分割结果呢？

ZhaochongAn commented 2 months ago

Hi, our visualization code renders all the results one by one. If you follow the code, you will get all the results including the predictions. You could press 'q' to finish the current rendering window and continue to the next result automatically.

lrnb12345 commented 1 month ago

您好，我们的可视化代码会逐一渲染所有结果。如果您按照代码操作，您将获得所有结果（包括预测）。您可以按“q”完成当前渲染窗口并自动继续下一个结果。

Hello, I visualized the query prediction results, but the entire scene is black. Why is this ![Uploading 屏幕截图 2024-08-18 145037.png…]()

lrnb12345 commented 1 month ago

捕获 你好，这是可视化的结果，如何才能获得论文上那种特定颜色的分割结果呢？

Hello, may I ask if you have encountered a situation where the visualization of your prediction results is all black

lrnb12345 commented 2 weeks ago

捕获 你好，这是可视化的结果，如何才能获得论文上那种特定颜色的分割结果呢？

Hello, have you successfully saved the result as ply and visualized it? How can I do it