How long should homography take? #25
-
|
Any recommendations for speeding up the process? |
Beta Was this translation helpful? Give feedback.
Replies: 2 comments 3 replies
-
|
That depends on the computer specs. It should be using GPU by default if the CUDA libraries are correctly configured. When running on a CPU only system, there is multi-threading support provided but the speed in that case is still dependent on the CPU architecture. |
Beta Was this translation helpful? Give feedback.
-
|
Might have to move to HPC, LOL. |
Beta Was this translation helpful? Give feedback.
-
|
So after 24 hours it is still running :( What kind of computer set-up would you recommend to process these? I do not think something like this can be run on HPC. |
Beta Was this translation helpful? Give feedback.
-
|
Yes, running it on a single thread using CPU will take a while. It could totally run on HPC. The SocialEyes modules are meant to be independent and could be run individually using their main.py script. Only the glassesRecord module has the docker requirement, the others should be platform independent. |
Beta Was this translation helpful? Give feedback.
-
|
Just got it running on the cluster with cuda and wow is it fast (3 hours!). |
Beta Was this translation helpful? Give feedback.
Yes, running it on a single thread using CPU will take a while.
It could totally run on HPC. The SocialEyes modules are meant to be independent and could be run individually using their main.py script. Only the glassesRecord module has the docker requirement, the others should be platform independent.