search

JonasSchult / Mask3D

Mask3D predicts accurate 3D semantic instances achieving state-of-the-art on ScanNet, ScanNet200, S3DIS and STPLS3D.
MIT License
525 stars 103 forks source link

Confusion when loading visualizations. #140

Open LinLin1031 opened 10 months ago

LinLin1031 commented 10 months ago

I have set save_visualization=true to get the visualization of the instance segmentation. But I find that when I drag the generated .bin file saved in the visualization folder into CloudCompare, it shows so many points. You know that there are only a few hundred thousand points in my source data. May I ask what is going on here? How can I keep the visualization points consistent with the source data? @JonasSchult

The following figure shows the number of points in the bin file of the visualization. image

The following figure shows the number of points of the source data. image

JonasSchult commented 10 months ago

We use PyViz3D to visualize pointclouds in a web browser of your choice ( see here for an example).

You can start an http server in the visualization folder as follows: python -m http.server 6035

More details about PyViz3D.

Let me know if this helps. :)

Best, Jonas

windfill commented 10 months ago

Hi,can I take a look at your results and share them with me? @LinLin1031

LinLin1031 commented 10 months ago

Sorry, I don't have results on S3DIS at the moment. I used a custom dataset for my experiments. @windfill

windfill commented 10 months ago

Hi,custom dataset is OK,I already have results on S3DIS. I want to confirm my experimental results.Can I give you your QQ or other contact information?@LinLin1031

LinLin1031 commented 10 months ago

I'm not sure whether I can help you, my results on the custom dataset are not quite satisfactory yet, and I'm still continuing to train. My QQ is 983504527. @windfill