Created release v19.22
dlib/CMakeLists.txt | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
Add matrix pointwise_pow (#2329)
dlib/matrix/matrix_utilities.h | 39 +++++++++++++++++++++++++++++++++ dlib/matrix/matrix_utilities_abstract.h | 21 ++++++++++++++++++ dlib/test/matrix4.cpp | 14 ++++++++++++ 3 files changed, 74 insertions(+)
Add softmax function for matrix type (#2320)
dlib/matrix/matrix_math_functions.h | 34 ++++++++++++++++++++++++++++ dlib/matrix/matrix_math_functions_abstract.h | 16 +++++++++++++ dlib/test/matrix.cpp | 17 ++++++++++++++ 3 files changed, 67 insertions(+)
fix invalid assert
dlib/global_optimization/global_function_search.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
[SERIALIZATION] updated _abstract file (#2306)
dlib/vectorstream/vectorstream_abstract.h | 22 ++++++++++++++-------- 1 file changed, 14 insertions(+), 8 deletions(-)
Serialization to and from vector<int8_t> and vector<uint8_t> (#2301)
dlib/serialize.h | 38 ++++++++++++++++++++++++ dlib/test/dnn.cpp | 5 ++++ dlib/test/matrix.cpp | 26 +++++++++++++++-- dlib/test/serialize.cpp | 63 ++++++++++++++++++++++++++++++++++++++++ dlib/test/vectorstream.cpp | 34 +++++++++++++++++++--- dlib/vectorstream/vectorstream.h | 49 ++++++++++++++++++++++++------- 6 files changed, 199 insertions(+), 16 deletions(-)
fix set_learning_rate_multipliers_range not working (#2304)
dlib/dnn/utilities.h | 4 ++-- dlib/test/dnn.cpp | 22 +++++++++++++++++++++- 2 files changed, 23 insertions(+), 3 deletions(-)
Minor fix in the network size format (#2303)
dlib/dnn/trainer.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Add missing .get_final_data_gradient() for repeat layer
dlib/dnn/core.h | 3 +++ 1 file changed, 3 insertions(+)
[MATRIX] bug fix. If you #include <dlib/matrix.h> in a header file you might get a compiler error saying ambiguous call to max(). This commit fixes that. (#2287)
dlib/pixel.h | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-)
work around bug in visual studio
dlib/test/global_optimization.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
FFT: added std::vector overloads for fft, ifft, fft_inplace and ifft_inplace (#2286)
dlib/matrix/matrix_fft.h | 75 ++++++++++++++++++++++++++++++++++++++- dlib/matrix/matrix_fft_abstract.h | 69 +++++++++++++++++++++++++++++++++++ dlib/test/fft.cpp | 74 ++++++++++++++++++++++++++++++++++++++ 3 files changed, 217 insertions(+), 1 deletion(-)
I'm finding there are other issues with gcc 4.8 now and it's old enough that we can just drop support for gcc 4.8.
dlib/global_optimization/find_max_global.h | 21 +-------------------- 1 file changed, 1 insertion(+), 20 deletions(-)
Early termination for find_{min,max}_global (#2281)
dlib/global_optimization/find_max_global.h | 16 ++++++--- .../global_optimization/find_max_global_abstract.h | 40 +++++++++++++++++----- dlib/test/global_optimization.cpp | 18 +++++++--- 3 files changed, 58 insertions(+), 16 deletions(-)
work around bug in gcc 4.8
dlib/global_optimization/find_max_global.h | 21 ++++++++++++++++++++- dlib/test/global_optimization.cpp | 10 ++++++++++ 2 files changed, 30 insertions(+), 1 deletion(-)
Greatly simplify how all the overloads of find_*_global() are setup.
dlib/global_optimization/find_max_global.h | 670 ++++++--------------- .../global_optimization/find_max_global_abstract.h | 540 +---------------- dlib/test/global_optimization.cpp | 13 +- 3 files changed, 190 insertions(+), 1033 deletions(-)
Fix compiler warning
dlib/cuda/cuda_dlib.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
fix find_min_global finding wrong answers (#2284)
dlib/global_optimization/find_max_global.h | 5 +++++ 1 file changed, 5 insertions(+)
Fix build with clang on ppc64le (#2278)
dlib/simd/simd4f.h | 5 +++++ 1 file changed, 5 insertions(+)
Fix build errors in gcc 4.8
dlib/test/fft.cpp | 3 +-- dlib/test/fftr_good_data.h | 4 ++-- 2 files changed, 3 insertions(+), 4 deletions(-)
Arbitrary sized FFTs using modified kissFFT as default backend and MKL otherwise (#2253)
dlib/cmake_utils/find_blas.cmake | 3 + dlib/matrix/fft_size.h | 220 +++ dlib/matrix/kiss_fft.h | 746 +++++++++ dlib/matrix/matrix_fft.h | 919 ++--------- dlib/matrix/matrix_fft_abstract.h | 75 +- dlib/matrix/matrix_utilities.h | 34 + dlib/matrix/mkl_fft.h | 243 +++ dlib/test/fft.cpp | 443 +++++- dlib/test/fftr_good_data.h | 3044 +++++++++++++++++++++++++++++++++++++ 9 files changed, 4880 insertions(+), 847 deletions(-)
Remove GUI dependency from fonts (#2273)
dlib/CMakeLists.txt | 2 +- dlib/gui_widgets/fonts.cpp | 2 ++ dlib/gui_widgets/fonts.h | 9 ++++----- dlib/gui_widgets/fonts_abstract.h | 8 +++++--- dlib/gui_widgets/nativefont.h | 5 ++++- dlib/image_transforms/draw.h | 2 +- 6 files changed, 17 insertions(+), 11 deletions(-)
Draw string on images (#2270)
dlib/image_transforms/draw.h | 99 +++++++++++++++++++++++++++++++++++ dlib/image_transforms/draw_abstract.h | 30 +++++++++++ dlib/test/image.cpp | 30 ++++++++++- 3 files changed, 158 insertions(+), 1 deletion(-)
Fixed CUDA 11.1 compilation. (#2271)
dlib/cuda/cuda_dlib.cu | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-)
make test run faster
dlib/test/rand.cpp | 25 +++++++++++-------------- 1 file changed, 11 insertions(+), 14 deletions(-)
Make the mmod loss not spam messages in the unit tests
dlib/dnn/loss.h | 27 ++++++++++++++++++--------- dlib/dnn/loss_abstract.h | 5 +++++ dlib/test/dnn.cpp | 1 + 3 files changed, 24 insertions(+), 9 deletions(-)
Fix build error from ambiguous calls to call_if_valid()
dlib/metaprogramming.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
Release GIL before detecting faces (#2252)
tools/python/src/simple_object_detector_py.h | 18 ++++++++++++++---- 1 file changed, 14 insertions(+), 4 deletions(-)
Update dnn_inception_ex.cpp (#2256)
examples/dnn_inception_ex.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Removed redundant calls that сause compilation nodiscard warnings. (#2251)
dlib/bigint/bigint_kernel_1.h | 5 ++--- dlib/bigint/bigint_kernel_2.h | 5 ++--- dlib/bigint/bigint_kernel_c.h | 5 ++--- 3 files changed, 6 insertions(+), 9 deletions(-)
Some systems manage to install cuda in a way that causes cmake to claim that cuda is available even though it knows it didn't find cublas, which is part of the standard cuda install. So we need to add a check to see if cmake *really for realz* found cuda.
dlib/CMakeLists.txt | 6 ++++++ 1 file changed, 6 insertions(+)
Adding exponential and weibull distributions (#2247)
dlib/rand/rand_kernel_1.h | 25 +++++++++++++++++++++++ dlib/rand/rand_kernel_abstract.h | 21 +++++++++++++++++++ dlib/test/rand.cpp | 44 ++++++++++++++++++++++++++++++++++++++++ 3 files changed, 90 insertions(+)
minor cleanup
dlib/data_io/cifar.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Add CIFAR-10 dataset loader (#2245)
dlib/CMakeLists.txt | 1 + dlib/all/source.cpp | 1 + dlib/data_io.h | 1 + dlib/data_io/cifar.cpp | 106 ++++++++++++++++++++++++++++++++++++++++++ dlib/data_io/cifar.h | 31 ++++++++++++ dlib/data_io/cifar_abstract.h | 51 ++++++++++++++++++++ dlib/data_io/mnist.cpp | 8 ++-- 7 files changed, 195 insertions(+), 4 deletions(-)
Fixes bug when (de)serializing vector<complex<float>> (#2244)
dlib/serialize.h | 106 ++++++++++++++++++++++++------------------------ dlib/test/serialize.cpp | 11 +++-- 2 files changed, 61 insertions(+), 56 deletions(-)
Rename function to disable_duplicative_biases (#2246)
dlib/dnn/layers.h | 2 +- dlib/dnn/layers_abstract.h | 2 +- dlib/test/dnn.cpp | 6 +++--- examples/dnn_dcgan_train_ex.cpp | 4 ++-- 4 files changed, 7 insertions(+), 7 deletions(-)
Add support for matrix serialization to python API (#2241)
tools/python/src/matrix.cpp | 12 ++++++++++++ 1 file changed, 12 insertions(+)
Expose get_face_chip_details to Python (#2238)
tools/python/src/image2.cpp | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+)
fix unused parameter warning in visitor_net_to_xml (#2240)
dlib/dnn/utilities.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
Make dnn_trainer print the minibatch size to ostream (#2236)
dlib/dnn/trainer.h | 1 + 1 file changed, 1 insertion(+)
Add custom ostream to console progress indicator (#2234)
dlib/console_progress_indicator.h | 36 +++++++++++++++++++----------------- 1 file changed, 19 insertions(+), 17 deletions(-)
Update to pybind11 v2.2.4 (closes #1806) (#2229)
dlib/external/pybind11/CMakeLists.txt | 2 + dlib/external/pybind11/README.md | 2 +- dlib/external/pybind11/include/pybind11/cast.h | 14 +++--- .../pybind11/include/pybind11/detail/common.h | 15 ++++--- .../pybind11/include/pybind11/detail/internals.h | 44 +++++++++++++++++-- dlib/external/pybind11/include/pybind11/eigen.h | 8 ---- dlib/external/pybind11/include/pybind11/iostream.h | 2 +- dlib/external/pybind11/include/pybind11/numpy.h | 7 +-- dlib/external/pybind11/include/pybind11/pybind11.h | 50 +++++++++++----------- dlib/external/pybind11/include/pybind11/stl.h | 10 ++++- dlib/external/pybind11/tools/pybind11Tools.cmake | 7 +++ 11 files changed, 106 insertions(+), 55 deletions(-)
slightly improve tests
dlib/test/pixel.cpp | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-)
Fix pixels being rounded to int values in some cases (#2228)
dlib/matrix/matrix_utilities.h | 2 +- dlib/test/image.cpp | 17 +++++++++++++++++ dlib/test/pixel.cpp | 8 +++++--- 3 files changed, 23 insertions(+), 4 deletions(-)
Remove an unused variable and old commented code (#2217)
dlib/cuda/cuda_dlib.cu | 3 --- 1 file changed, 3 deletions(-)
Add Layer Normalization (#2213)
dlib/cuda/cpu_dlib.cpp | 171 +++++++++++++++++++++++++++++++++++++ dlib/cuda/cpu_dlib.h | 24 ++++++ dlib/cuda/cuda_dlib.cu | 163 +++++++++++++++++++++++++++++++++++ dlib/cuda/cuda_dlib.h | 26 +++++- dlib/cuda/tensor_tools.cpp | 34 ++++++++ dlib/cuda/tensor_tools.h | 24 ++++++ dlib/dnn/layers.h | 185 ++++++++++++++++++++++++++++++++++++++-- dlib/dnn/layers_abstract.h | 149 +++++++++++++++++++++++++++++++- dlib/test/dnn.cpp | 102 ++++++++++++++++++++++ examples/dnn_dcgan_train_ex.cpp | 4 +- 10 files changed, 869 insertions(+), 13 deletions(-)
fix backtracking when losses stay at inf (fixes #2206) (#2209)
dlib/dnn/trainer.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
Do not use sqrt_2 in device code (fixes #2208) (#2210)
dlib/cuda/cpu_dlib.cpp | 2 +- dlib/cuda/cuda_dlib.cu | 3 ++- 2 files changed, 3 insertions(+), 2 deletions(-)
Add GELU activation layer (#2204)
dlib/cuda/cpu_dlib.cpp | 41 ++++++++++++++++++++++++++ dlib/cuda/cpu_dlib.h | 13 +++++++++ dlib/cuda/cuda_dlib.cu | 54 ++++++++++++++++++++++++++++++++++ dlib/cuda/cuda_dlib.h | 13 +++++++++ dlib/cuda/tensor_tools.cpp | 27 +++++++++++++++++ dlib/cuda/tensor_tools.h | 34 ++++++++++++++++++++++ dlib/dnn/layers.h | 72 ++++++++++++++++++++++++++++++++++++++++++++++ dlib/dnn/layers_abstract.h | 38 ++++++++++++++++++++++++ dlib/test/dnn.cpp | 33 +++++++++++++++++++++ 9 files changed, 325 insertions(+)
fix cmake not finding openblas
dlib/cmake_utils/find_blas.cmake | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-)
Test cuda losses (#2199)
dlib/cuda/cpu_dlib.h | 142 +++++++++++++++++++++++++++++++++++++----- dlib/dnn/loss.h | 130 ++++---------------------------------- dlib/dnn/misc.h | 43 +++++++++++++ dlib/dnn/utilities.h | 17 ----- dlib/dnn/utilities_abstract.h | 2 + dlib/test/dnn.cpp | 56 +++++++++++++++-- 6 files changed, 235 insertions(+), 155 deletions(-)
Add loss multiclass log per pixel weighted cuda (#2194)
dlib/cuda/cpu_dlib.h | 77 ++++++++++++++++++++++++++++++++++++++++++++++++ dlib/cuda/cuda_dlib.cu | 48 ++++++++++++++++++++++++++++++ dlib/cuda/cuda_dlib.h | 77 ++++++++++++++++++++++++++++++++++++++++++++++++ dlib/dnn/loss.h | 63 ++++++++------------------------------- dlib/dnn/loss_abstract.h | 3 +- dlib/dnn/misc.h | 28 ++++++++++++++++++ dlib/test/dnn.cpp | 12 ++++++++ 7 files changed, 256 insertions(+), 52 deletions(-)
DLIB (de)serialization : enhanced STL container support (#2185)
dlib/serialize.h | 660 +++++++++++++++++++++++++++++++++++++++++++++++- dlib/test/serialize.cpp | 91 ++++++- 2 files changed, 737 insertions(+), 14 deletions(-)
Linux Distro Detection to fix issue number #2159 #154 (#2169)
dlib/cmake_utils/find_blas.cmake | 13 +++++++------ 1 file changed, 7 insertions(+), 6 deletions(-)
Let python users give up to 35 parameters when using the global optimizer.
tools/python/src/global_optimization.cpp | 22 +++++++++++++++++++++- 1 file changed, 21 insertions(+), 1 deletion(-)
update docs
dlib/serialize.h | 10 ++++++++++ 1 file changed, 10 insertions(+)
Extended proxy_(de)serialize objects to work with stringstream, ostringstream, istringstream and vector<char> (#2181)
dlib/serialize.h | 111 ++++++++++++++++++++++++++++++++++-------------- dlib/test/serialize.cpp | 47 +++++++++++++++++--- 2 files changed, 119 insertions(+), 39 deletions(-)
use DLIB_CASSERT to avoid unused variable warning in release compilation (#2182)
dlib/vectorstream/vectorstream.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
vectorstream updates: added seekoff and seekpos (#2179)
dlib/test/vectorstream.cpp | 81 ++++++++++++++++++++++++++++++++++------ dlib/vectorstream/vectorstream.h | 41 +++++++++++++++----- 2 files changed, 101 insertions(+), 21 deletions(-)
add some docs
dlib/serialize.h | 102 +++++++++++++++++++++++++++++++++++++++---------------- 1 file changed, 73 insertions(+), 29 deletions(-)
Macro for generating default serialisation functions (#2177)
dlib/serialize.h | 72 ++++++++++++++++++++++++++++++++++++++++++++++++- dlib/test/serialize.cpp | 50 ++++++++++++++++++++++++++++++++++ 2 files changed, 121 insertions(+), 1 deletion(-)
Add scale_prev layer (#2171)
dlib/dnn/layers.h | 122 +++++++++++++++++++++++++++++++++++++++++++++ dlib/dnn/layers_abstract.h | 74 +++++++++++++++++++++++++++ dlib/test/dnn.cpp | 53 ++++++++++++++++++++ 3 files changed, 249 insertions(+)
Add error message for mismatched tensor sizes in dnn_trainer (#2165)
dlib/dnn/trainer.h | 5 +++++ 1 file changed, 5 insertions(+)
Simplified more uses of layer visiting and fixed constness bug
dlib/dnn/core.h | 24 ++++++++++++++++---- dlib/dnn/utilities.h | 36 +++++------------------------- dlib/test/dnn.cpp | 62 ++++++++++++++++++++++++++++++++++++++++++---------- 3 files changed, 76 insertions(+), 46 deletions(-)
Show how to use the new visitors with lambdas (#2162)
examples/dnn_dcgan_train_ex.cpp | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
switch this to C++11 code
dlib/test/metaprogramming.cpp | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
make type_safe_union movable and also support holding movable types in a natural way.
dlib/test/type_safe_union.cpp | 36 ++++++++++++++++++++++ dlib/type_safe_union/type_safe_union_kernel.h | 27 +++++++++++----- .../type_safe_union_kernel_abstract.h | 25 ++++++++++++--- 3 files changed, 76 insertions(+), 12 deletions(-)
Made the DNN layer visiting routines more convenient.
dlib/dnn/core.h | 279 +++++++++++++++++++------------------- dlib/dnn/core_abstract.h | 81 ++++++++++- dlib/dnn/trainer.h | 2 +- dlib/test/dnn.cpp | 16 ++- examples/dnn_introduction3_ex.cpp | 22 ++- 5 files changed, 236 insertions(+), 164 deletions(-)
Added call_if_valid()
dlib/metaprogramming.h | 40 ++++++++++++++++++++++++++++++++++++++++ dlib/test/metaprogramming.cpp | 37 +++++++++++++++++++++++++++++++++++++ 2 files changed, 77 insertions(+)
Add visitor to remove bias from bn_ layer inputs (#closes 2155) (#2156)
dlib/dnn/core.h | 22 ++++ dlib/dnn/core_abstract.h | 14 +++ dlib/dnn/layers.h | 224 ++++++++++++++++++++++++++++++---------- dlib/dnn/layers_abstract.h | 64 ++++++++++++ examples/dnn_dcgan_train_ex.cpp | 43 +++----- 5 files changed, 282 insertions(+), 85 deletions(-)
Make dnn_trainer use robust statistic to determine if the loss is exploding and if it should backtrack.
dlib/dnn/trainer.h | 16 ++++------------ 1 file changed, 4 insertions(+), 12 deletions(-)
dnn_trainer prints number of steps executed when print to ostream
dlib/dnn/trainer.h | 1 + 1 file changed, 1 insertion(+)
Add probability_values_are_increasing() and probability_values_are_increasing_robust()
dlib/statistics/running_gradient.h | 44 +++++++++++++++++++++++++++++ dlib/statistics/running_gradient_abstract.h | 39 +++++++++++++++++++++++++ dlib/test/statistics.cpp | 10 +++++++ 3 files changed, 93 insertions(+)
Rename POSIX macro to DLIB_POSIX to avoid name clashes with some libraries.
dlib/dir_nav/dir_nav_kernel_2.cpp | 4 ++-- dlib/gui_core/gui_core_kernel_2.cpp | 4 ++-- dlib/gui_widgets/nativefont.h | 4 ++-- dlib/misc_api/misc_api_kernel_2.cpp | 4 ++-- dlib/platform.h | 8 ++++---- dlib/sockets/sockets_kernel_2.cpp | 4 ++-- dlib/threads/threads_kernel_2.cpp | 4 ++-- 7 files changed, 16 insertions(+), 16 deletions(-)
Use the box with bounding box regression applied to do NMS in the loss.
dlib/dnn/loss.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
Fix bug in bounding box regression loss.
dlib/dnn/loss.h | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
Add missing input/output mappings to mult_prev (#2154)
dlib/dnn/layers.h | 3 +++ dlib/dnn/layers_abstract.h | 2 ++ 2 files changed, 5 insertions(+)
Fix a warning and add some more error handling.
dlib/dnn/loss.h | 5 +++++ dlib/dnn/loss_abstract.h | 5 +++-- dlib/test/dnn.cpp | 6 ++++-- 3 files changed, 12 insertions(+), 4 deletions(-)
loss multibinary log (#2141)
dlib/dnn/loss.h | 118 +++++++++++++++++++++++++++++++++++++++++++++++ dlib/dnn/loss_abstract.h | 83 +++++++++++++++++++++++++++++++++ dlib/test/dnn.cpp | 64 +++++++++++++++++++++++++ 3 files changed, 265 insertions(+)
Problem: With certain batch size / device count combinations, batches were generated with size = 1, causing problems when using batch normalization. (#2152)
dlib/dnn/trainer.h | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-)
switch to a name less likely to conflict with third party code
dlib/assert.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Minor typo fixes (#2150)
dlib/cuda/cudnn_dlibapi.cpp | 24 ++++++++++++------------ 1 file changed, 12 insertions(+), 12 deletions(-)
Fix warning for zero variadic macro arguments. (#2151)
dlib/assert.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Added "get_random_complex_gaussian" to dlib::rand (#2149)
dlib/rand/rand_kernel_1.h | 28 ++++++++++++++++++---------- dlib/rand/rand_kernel_abstract.h | 9 +++++++++ 2 files changed, 27 insertions(+), 10 deletions(-)
fix typo, doesn't really matter, but still
dlib/global_optimization/find_max_global.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Record last changeset and set PATCH version to 99
dlib/CMakeLists.txt | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Created release v19.21
dlib/CMakeLists.txt | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)