Closed ffiioonnaa closed 4 years ago
It looks like the Nvidia OpenGL libraries are missing -- there should be files like libEGL_nvidia.so.0
in /usr/lib/x86_64-linux-gnu
. I think if you reinstall the nvidia driver using the Ubuntu package version, not the nvidia runfile, then it will fix this. Otherwise, see if the libraries are somewhere non-standard, by doing find / -name *GL*nvidia*
-- which will search the whole system for them.
It works! Thanks for replying so quickly!
And another question, I want to render a batch of meshes after reconstruct them, and then to calculate the loss with the rendered images, can dirt do that? and the mesh has several part not just one manifold. thanks.
Yes, DIRT will render a batch of meshes with rasterise_batch
, then you calculate the loss in the usual way (e.g. mean square difference) on the resulting images in tensorflow. DIRT will ensure the correct gradient backpropagates through the rendering. Yes, non-manifold meshes are ok. If you have other problems/questions, please open a new ticket!
Hi: Thanks for your excellent work! After 'pip install . ' I encountered an error: none of 2 egl devices matches the active cuda device. I 've seen other issues about this error ,but it seems the CMakeCache.txt has changed, and I'am not sure how to change it. Can you help me see what the problem is? My configuration: Ubuntu 18.04 pthyon3.6 tensorflow1.13.1 CUDA10.0 nvidia driver 430
ldd dirt/librasterise.so
and I find nvidia-430 folder in /usr/src rather than /usr/lib. ls -l /usr/lib//GL*
Thanks again. I will be appreciated for your reply as soon as possible, thank you very much!