Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Problem with model not being able to view #51

Open
xdw-123 opened this issue May 28, 2024 · 21 comments
Open

Problem with model not being able to view #51

xdw-123 opened this issue May 28, 2024 · 21 comments

Comments

@xdw-123
Copy link

xdw-123 commented May 28, 2024

Thank you for the author's sharing. I tried to achieve the effect of your paper using my own data. I modified the data input parameters of train_db. sh for reconstruction, but encountered the following problem when opening the model later. Hope for your answer
image
image

@tongji-rkr
Copy link
Collaborator

We update the viewer and you can try it again

@tongji-rkr
Copy link
Collaborator

Please don't put the .pt file under the current path, just keep the original file structure after training

@xdw-123
Copy link
Author

xdw-123 commented May 28, 2024

Thank you for the author's reply. I have updated the code and recompiled it to read. pt. Is the last file very important? The program still reported an error
image

@tongji-rkr
Copy link
Collaborator

No, this is an optional parameter.
Please show me your current file structure.

@xdw-123
Copy link
Author

xdw-123 commented May 28, 2024

image
image

@tongji-rkr
Copy link
Collaborator

No problem, and any bugs here?

@xdw-123
Copy link
Author

xdw-123 commented May 28, 2024

Because I previously implemented the Octree GS file you shared, and found out that the path to the. pt file was incorrect. Later, I corrected it, but found that the model had distortion in details. However, combining the LOD principle is a great practice. Seeing that your shared file has solved the distortion problem, I am curious to implement it and see the effect.

@xdw-123
Copy link
Author

xdw-123 commented May 28, 2024

Sorry to throw the compilation issue directly at you, mainly because I am in a hurry to see the results. Would it be convenient for you to share the execution program of your viewer?

@tongji-rkr
Copy link
Collaborator

tongji-rkr commented May 28, 2024

I get what you meant, actually I already gave a pre-compiled version, but a lot of people don't read ReadME
https://drive.google.com/file/d/17nPVnRRxO4zMQJ_fKHm13HzvD0wUZ8XF/view?usp=sharing

@xdw-123
Copy link
Author

xdw-123 commented May 28, 2024

Excuse me, is this the compiler that was just compiled? If not, then I have tried downloading and won't be able to read the. pt file

@tongji-rkr
Copy link
Collaborator

The following is a compiler error, I am using VS2019 image
I configured it in win11+vs2022 and I never encountered any errors related to it

@tongji-rkr
Copy link
Collaborator

Thank you for the author's reply. I have updated the code and recompiled it to read. pt. Is the last file very important? The program still reported an error image

That seems fine. Would you like to Google your results to me.

@tongji-rkr
Copy link
Collaborator

Because I previously implemented the Octree GS file you shared, and found out that the path to the. pt file was incorrect. Later, I corrected it, but found that the model had distortion in details. However, combining the LOD principle is a great practice. Seeing that your shared file has solved the distortion problem, I am curious to implement it and see the effect.

We had some problems with the initial code, and later I added an anchor culling strategy, which you can refer to the weed_out function in gaussian_model.py

@xdw-123
Copy link
Author

xdw-123 commented May 28, 2024

Thank you for the author's reply. I have updated the code and recompiled it to read. pt. Is the last file very important? The program still reported an error image

That seems fine. Would you like to Google your results to me.

This is the viewer I compiled, not the viewer execution program you provided.I seem to be compiling Torch, but there seems to be an issue here. Maybe it doesn't support CUDA, right
image

@tongji-rkr
Copy link
Collaborator

Which version of libtorch are you using?

@xdw-123
Copy link
Author

xdw-123 commented May 28, 2024

Create a virtual environment using the conda you used in your tutorial, Does the Windows compilation viewer use the libtorch 1.10 version you shared, which is absolutely related to the cuda version?

@tongji-rkr
Copy link
Collaborator

tongji-rkr commented May 28, 2024

CUDA 11.6+ #32
CUDA 11.8 best

@xdw-123
Copy link
Author

xdw-123 commented May 28, 2024

I just tried the version of CUDA11.6, but it still reported an error in the same place. Maybe the version is still incorrect, I think Ubuntu also has CUDA 11.6

@tongji-rkr
Copy link
Collaborator

https://github.com/city-super/Scaffold-GS/tree/main/SIBR_viewers
You go through all these steps and get an error at compile time, right?

@tongji-rkr
Copy link
Collaborator

if so, please tell me all environment configuration

@xdw-123
Copy link
Author

xdw-123 commented May 29, 2024

Hello, yesterday I tested different versions of CUDA and also used 11.8, but still encountered an error when reading the PT file. Then I looked at the Torch version in the virtual environment again, and the image is shown below. I downloaded the LibTorch package for this version (cu116+1.12.1), but it still reported an error in the same place.
1716945790827

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants