search

ToughStoneX / Self-Supervised-MVS

Pytorch codes for "Self-supervised Multi-view Stereo via Effective Co-Segmentation and Data-Augmentation"
152 stars 15 forks source link

cannot build fusible #11

Open derrick-xwp opened 2 years ago

derrick-xwp commented 2 years ago

Dear author,

I cannot build fusbile. Can you send me an executable one?

ToughStoneX commented 2 years ago

The built one in my computer may not be able to run on your computer. Could you share the reported errors when building the fusible?

derrick-xwp commented 2 years ago

'roundf': definition of dllimport function not allowed

THanks

ToughStoneX commented 2 years ago

Are you building the code in Windows? It is mentioned that the provided code does not support Windows. At least I have never tried to build it in Windows. Please build the code in Linux.

derrick-xwp commented 2 years ago

Yes Will the performance degrade if run on windows?

derrick-xwp commented 2 years ago

Have you tried on Mac?

derrick-xwp commented 2 years ago

Dear author,

I also would like to confirm with you regarding the experiment configuration.

Are most experiments carried in following config:

resolution: 1200 X 1600, number of source view is 5, on test set scenes.

Thanks

ToughStoneX commented 2 years ago

1, I have not tried on Windows or Mac. The code is only tested on Linux server with GPUs. 2, Most experiments are using the default setting as you mentioned. However, due to the memory overflow in a 11G GPU like 1080Ti/2080Ti, some experiments use a lower resolution of 1152 x 864 with MVSNet.

derrick-xwp commented 2 years ago

Dear author, Is the bellow error thrown out when executing fusible a normal one?
Error: no kernel image is available for execution on the device?

Thanks!

ToughStoneX commented 2 years ago

Maybe the fusible is not built correctly. Here are some suggestions: 1, Check if OpenCV is correctly installed in your computer. 2, Check if the directory fed to the executable file of fusible is correct. (If you did not modify the original code, you can ignore this issue, the default setting would be fine.) 3, Check the GPU architecture of your server. Is it 1080Ti or 2080Ti? Modify the CMakeList.txt following the instruction in the README file: https://github.com/ToughStoneX/Self-Supervised-MVS/tree/main/jdacs#fusion, and rebuild the fusible executable file.