On fedora 30 using gcc9, ../core/libs/facesengine/recognition/dlib-dnn/dnnface/dnn_face.h:175:88: required from here ../core/libs/facesengine/recognition/dlib-dnn/dnnface/image_transforms/interpolation.h:1088:22: error: conversion from 'const double' to 'const simd4f' is ambiguous 1088 | const simd4f _tb_frac = tb_frac; | ^~~~~~~~ In file included from ../core/libs/facesengine/recognition/dlib-dnn/dnnface/image_transforms/interpolation.h:46, from ../core/libs/facesengine/facedb/facedb.cpp:44: ../core/libs/facesengine/recognition/dlib-dnn/dnnface/dnn_base/simd4f.h:128:16: note: candidate: 'simd4f::simd4f(float)' 128 | inline simd4f(float f) : x{f,f,f,f} { } | ^~~~~~ ../core/libs/facesengine/recognition/dlib-dnn/dnnface/dnn_base/simd4f.h:121:16: note: candidate: 'simd4f::simd4f(int)' 121 | inline simd4f(const vector float& v) : x{v} { } | ^~~~~~ In file included from ../core/libs/facesengine/facedb/facedb.cpp:44: ../core/libs/facesengine/recognition/dlib-dnn/dnnface/image_transforms/interpolation.h:1089:22: error: conversion from 'double' to 'const simd4f' is ambiguous 1089 | const simd4f _inv_tb_frac = 1-tb_frac; | ^~~~~~~~~~~~ In file included from ../core/libs/facesengine/recognition/dlib-dnn/dnnface/image_transforms/interpolation.h:46, from ../core/libs/facesengine/facedb/facedb.cpp:44: ../core/libs/facesengine/recognition/dlib-dnn/dnnface/dnn_base/simd4f.h:128:16: note: candidate: 'simd4f::simd4f(float)' 128 | inline simd4f(float f) : x{f,f,f,f} { } | ^~~~~~ ../core/libs/facesengine/recognition/dlib-dnn/dnnface/dnn_base/simd4f.h:121:16: note: candidate: 'simd4f::simd4f(int)' 121 | inline simd4f(const vector float& v) : x{v} { } | ^~~~~~ In file included from ../core/libs/facesengine/facedb/facedb.cpp:44: ../core/libs/facesengine/recognition/dlib-dnn/dnnface/image_transforms/interpolation.h:1090:22: error: conversion from 'double' to 'const simd4f' is ambiguous 1090 | const simd4f _x_scale = 4*x_scale; | ^~~~~~~~ In file included from ../core/libs/facesengine/recognition/dlib-dnn/dnnface/image_transforms/interpolation.h:46, from ../core/libs/facesengine/facedb/facedb.cpp:44: ../core/libs/facesengine/recognition/dlib-dnn/dnnface/dnn_base/simd4f.h:128:16: note: candidate: 'simd4f::simd4f(float)' 128 | inline simd4f(float f) : x{f,f,f,f} { } | ^~~~~~ ../core/libs/facesengine/recognition/dlib-dnn/dnnface/dnn_base/simd4f.h:121:16: note: candidate: 'simd4f::simd4f(int)' 121 | inline simd4f(const vector float& v) : x{v} { } | ^~~~~~ Full logs, https://koji.fedoraproject.org/koji/taskinfo?taskID=33066535 (interesting piece is build.log link)
*** Bug 405625 has been marked as a duplicate of this bug. ***
Maik, Typically, the current DNN code will be dropped in the future in favor to OpenCV DNN. No need to spare time on this topic... Gilles
So, does that mean we should turn off faces engine for our package builds until the OpenCV code comes?
There is no option in digiKam cmake scripts to disable whole face engine. The alternative is to add a cmake option to disable the DNN part which stll experimental. Note : we will have a student planed while this summer to port the DLib DNN to OpenCV DNN. So typically this code will disappear, but later this summer. Gilles Caulier
Git commit 6a3da298103797d01fa0cb049cccaff39bd1583e by Gilles Caulier. Committed on 21/03/2019 at 16:29. Pushed by cgilles into branch 'master'. add new optional configuration option to enable Faces Engine DNN support Related: bug 405625 M +2 -0 Mainpage.dox M +4 -1 NEWS M +1 -0 bootstrap.linux M +1 -0 bootstrap.macports M +1 -0 bootstrap.mxe M +1 -0 bootstrap.tarball M +8 -5 core/CMakeLists.txt M +6 -3 core/app/utils/digikam_config.h.cmake.in M +20 -11 core/libs/facesengine/CMakeLists.txt M +34 -22 core/libs/facesengine/facedb/facedb.cpp M +6 -0 core/libs/facesengine/facedb/facedb.h M +55 -8 core/libs/facesengine/recognitiondatabase.cpp M +1 -1 core/libs/facesengine/recognitiondatabase.h M +3 -0 core/utilities/facemanagement/facescandialog.cpp https://commits.kde.org/digikam/6a3da298103797d01fa0cb049cccaff39bd1583e
With 7.0.0, Dlib code have been dropped from digiKam face engine in favor of OpenCV Neural Network API. Gilles Caulier