aminebdj / openyolo3d Goto Github PK

Dear Authors,

How long does your method need to train?

Thanks

Hi Author,

I have few doubts regarding the testing dataset.

1. I am trying to download the Replica dataset but am unable to do so. How to do it? i am using
   sh scripts/get_class_agn_masks.sh.
2. Can you please explain how to generate the ground truth, poses, depth and point clouds (.ply) files? If possible, could you share the data?
3. Are you generating .ply through gaussian_splatting?
   These will be very helpful to test datasets to run with custom dataset.

Thank you in advance.

Dear authors,

Thank you for your awesome work and for releasing the code.

I found that at the run_evaluation.py, the text in the prediction = openyolo3d.predict() function is missing. I believe the text is necessary since as shown in the network architecture figure of your paper, the input contains the text, which plays an important role in making this method outstanding.

Could you please add the text into this function input, or explain this if the text is not needed in the evaluation process?

Thank you!

Dear authors,

Thanks for your fantastic work. It is very beneficial to the community of the 3D OV segmentation research field. May I ask if OpenYOLO3D enables segmentation with text prompts similar to SAM in 2D?

Thanks.

Best wishes,
Ye

Dear @aminebdj ,

Thank you again for your outstanding work! I have followed the instructions to reproduce the results. However, I encountered significant discrepancies when trying to reproduce the results for ScanNet200 using both precomputed masks and oracle 3D masks. The results I obtained are quite different from those reported in the paper. I have attached the results for both cases.

Could you please provide any insights or guidance on how to accurately reproduce the results?

Thank you very much!

Results of ScanNet200 with precomputed-masks:

Results of ScanNet200 with oracle 3D masks:

Hi authors,

Thank you for releasing the code for this great work!

I was able to successfully reproduce the performance in the paper.

But, I still wonder if there is any way that I can visualize the results.

Do you have any plan to release codes for visualizations?

or Would you give me a sample script / code that I can develop on for the visualizations?

Please feel free to contact through my email if you prefer email over github issue for sharing the code!

Email: shjung13 [at] cs.washington.edu

Thank you!

Best,
Sanghun

Hi,

Thank you so much for this great work! However, when I try to run the code, I noticed that the 'num_classes' parameter is missing in the config files for both the ScanNet200 and Replica datasets. This causes an error at line 256 in utils/init.py. Could you please let me know what values I should set for 'num_classes' for these datasets?

Thank you very much!