https://bugs.kde.org/show_bug.cgi?id=426175
--- Comment #79 from Marcel <maes...@gmail.com> --- So, I think it's time to summarize all findings: 1. Disable loading notification This solved the original problem 2. Huge memory demand at startup For this problem is another bug (https://bugs.kde.org/show_bug.cgi?id=426185). Some memory can be reduced anyway bug some is used to gain a speed advantage in digikam. The Question is what is more important. Speed or memory consumption. Personally, this memory for speed gain should definitely not be allocated at startup time as most digikam usages will not scan for faces each time they use digikam. It should be allocated at the start of face detection. 3. Memory leak on face detection There is a bug in openCV with the use of GPU causing that leak. More memory leaks on bigger picture collections. To solve that problem cv::Mat can be used instead of cv::UMat, but again that will decrease the scanning speed. Probably a checkbox "use GPU (experimental)" could help. -- You are receiving this mail because: You are watching all bug changes.