https://bugs.kde.org/show_bug.cgi?id=399923
--- Comment #175 from caulier.gil...@gmail.com --- Oh, i don't read well your last comment. It's interresting, especially this trace : > String, "autodetectedFace"), QVariant(int, 281), QVariant(int, 281), > QVariant(QString, "tagRegion")) > digikam.facedb: nnmodel: "" , nndata "" > digikam.facesengine: cv::Exception: OpenCV(3.4.7) > /build/source/modules/dnn/src/caffe/caffe_io.cpp:1121: error: (-2:Unspecified > error > ) FAILED: fs.is_open(). Can't open "" in function 'ReadProtoFromTextFile' > > digikam.general: Found 0 faces in "IMG-20170907-WA0003.jpg" QSize(1280, 720) > QSize(1280, 720) > digikam.facedb: nnmodel: "" , nndata "" > digikam.facesengine: cv::Exception: OpenCV(3.4.7) > /build/source/modules/dnn/src/caffe/caffe_io.cpp:1121: error: (-2:Unspecified > error > ) FAILED: fs.is_open(). Can't open "" in function 'ReadProtoFromTextFile' > > digikam.general: Found 0 faces in "IMG-20170907-WA0004.jpg" QSize(1032, 581) > QSize(1032, 581) > digikam.dimg: "/home/timo/tmp/deletable-pics/Syncthing-3TB/Oneplus > One/Whatsapp/Media/WhatsApp Images/IMG-20170907-WA0006.jpg" : JPE > G file identified It's clear : your OpenCV 3.4.7 is not well build. It miss symbols and generate a C++ exception. I build myself OpenCV 3.4.7, using OpenCV git repository and tag "3.4.7". I use a minimal configuration with this shell script : https://cgit.kde.org/digikam.git/tree/project/scripts/bootstrap.opencv-minimal-linux.sh Checkout opencv 3.4.7 tag from git repository, copy this shell script on root OpenCV directory, run it and go to build subdir to run make -j4 (as with digiKam source). Before to install this new minimal OpenCV, uninstall all previous development versions from OpenCV. Do not mix configurations. I already faced this kind of situation, and this make an unusable OpenCV with digiKam. You can let's OpenCv2 binaries as shared libraries, but all cmake config files, headers, and pkgconfig files must be removed and replaced by new compiled OpenCV version. Take a care to not use in parallel OpenCv4 files. Some part are binary compatible with OpenCv3 and OpenCv2. When this minimal OpenCv 3.4.7 is installed, from digiKam face recognition branch, remove build sub-dir, and reconfigure with the bootstrap script. Look while configuration stage is all OpenCv 3.4.7 modules are found properly. Compile and install this digiKam. Test again the scan for faces function. Gilles Caulier PS : here i test with Mageia6 which only includes OpenCv2. I uninstalled all OPenCv2 devel packages and checkout/compile/install OpenCv 3.4.7 as well. Mageia6 is installed on a real computer, not a VM. I use a NVIDIA video Card, even if my minimal OpenCv 3.4.7 is compiled without CUDA support. In fact i disabled all video and other system accelerations: it's not optimum but safe to run. -- You are receiving this mail because: You are watching all bug changes.
