https://bugs.kde.org/show_bug.cgi?id=522887

--- Comment #3 from [email protected] ---
@Maik

OpenCV5 use a new DNN module with the following changes :

- Caffe and Darknet model formats support are dropped. We must use ONNX format
everywhere. It still one .pb model used in digiKam for the aesthetic detector.
It must be converted, but this is not the problem for the face engine, as all
other models are ONNX.
- The new DNN module cover 80% of the ONNX operators (22% before), but the
module is CPU only for the moment. We need to switch to the previous module
with the rule "export OPENCV_FORCE_DNN_ENGINE=1" before to run digiKam
- OpenCV5 require C++17 everywhere in source code using these API, else
run-time errors can happen. I think it's already the case.

@th.brun

Please try to run digiKam like this in a console to see if it's better:

export OPENCV_FORCE_DNN_ENGINE=1 && digikam

Gilles Caulier

-- 
You are receiving this mail because:
You are watching all bug changes.

Reply via email to