Hi, Is it possible to get Digikam face recognition working with GPU (Cuda)? Digikam uses OpenCV and this can be accelerated using GPU on Linux (probably more complicated on Windows). I tried to compile version 8 and on the system ( Arch - Garuda ) I had a version of OpenCV installed with Cuda support. In the Digikam settings, I unchecked the option not to use the GPU, but still the face recognition does not use the GPU.
1 Like
This is a great suggestion. Yolo / OpenCV can be accelerated using CUDA, so with the right code in Digikam it should be possible to have the option to use GPU instead of CPU for face detection and recognition. (I am not familiar enough with the frameworks to know if you can use both CPU and GPU.)
Is there a formal way to suggest this as a future feature?
Instead of CUDA (NVidia-specific), OpenCL might be a better choice for GPU use, as it is supported on more GPUs, notably AMD. OpenCV supports both, not sure about YOLO.
In both cases, specific code is required to test for the presence of a usable GPU and to communicate with it.
The best way forward would be a feature request in the bug tracker (note that you need to have or create an account on the bugtracker)