Created release v20.1
dlib/CMakeLists.txt | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
adjusts version
dlib/CMakeLists.txt | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
Add text_classification_ex.cpp (#3182)
dlib/statistics/feature_hashing.h | 147 +++++++++--- dlib/statistics/feature_hashing_abstract.h | 67 ++++-- dlib/svm/auto.h | 146 ++++++++++++ dlib/svm/auto_abstract.h | 42 ++++ dlib/test/CMakeLists.txt | 1 + dlib/test/auto_train_linear_classifier.cpp | 197 ++++++++++++++++ dlib/test/feature_hashing.cpp | 32 +++ dlib/test/makefile | 1 + examples/CMakeLists.txt | 1 + examples/text_classification_ex.cpp | 357 +++++++++++++++++++++++++++++ 10 files changed, 945 insertions(+), 46 deletions(-)
Merge pull request #3184 from arrufat/aligned-union
Preserve type_safe_union storage layout
dlib/test/type_safe_union.cpp | 58 +++++++++++++++++++++++++++ dlib/type_safe_union/type_safe_union_kernel.h | 8 +++- 2 files changed, 65 insertions(+), 1 deletion(-)
A bit more checking/testing
dlib/svm/svm.h | 5 +++++ dlib/test/probabilistic.cpp | 19 +++++++++++++++++-- 2 files changed, 22 insertions(+), 2 deletions(-)
Replace the deprecated std::aligned_union in type_safe_union
dlib/type_safe_union/type_safe_union_kernel.h | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
Require each Platt fold to hold out both classes
dlib/svm/svm.h | 12 +++++++----- dlib/svm/svm_abstract.h | 8 ++++++-- dlib/test/probabilistic.cpp | 16 ++++++++++++++++ 3 files changed, 29 insertions(+), 7 deletions(-)
Add hash_word_ngrams() and hash_char_ngrams() (#3181)
dlib/statistics.h | 1 + dlib/statistics/feature_hashing.h | 140 +++++++++++++++++++++++++ dlib/statistics/feature_hashing_abstract.h | 81 +++++++++++++++ dlib/test/CMakeLists.txt | 1 + dlib/test/feature_hashing.cpp | 158 +++++++++++++++++++++++++++++ dlib/test/makefile | 1 + 6 files changed, 382 insertions(+)
Pass unsigned char to the <cctype> functions (#3179)
dlib/cpp_tokenizer/cpp_tokenizer_kernel_1.h | 2 +- dlib/data_io/arc_agi.h | 8 ++++---- dlib/gui_widgets/widgets.cpp | 8 ++++---- dlib/server/server_http.h | 4 ++-- dlib/string/string.h | 12 ++++++------ dlib/test/string.cpp | 7 +++++++ 6 files changed, 24 insertions(+), 17 deletions(-)
prepare for new release, basically just updating docs
dlib/CMakeLists.txt | 4 +- .../cpp_pretty_printer_kernel_1.h | 25 +++- .../cpp_pretty_printer_kernel_2.h | 9 ++ .../cpp_pretty_printer_kernel_abstract.h | 16 +++ dlib/dnn/onnx_abstract.h | 44 ++++--- dlib/geometry/line_abstract.h | 2 +- .../full_object_detection_abstract.h | 9 +- dlib/sockets/sockets_kernel_abstract.h | 11 +- dlib/tokenizer/bpe_tokenizer_abstract.h | 8 +- dlib/unicode/unicode_abstract.h | 6 +- tools/htmlify/htmlify.cpp | 133 +++++++++++++++++++-- tools/htmlify/test_htmlify.py | 118 ++++++++++++++++++ tools/python/src/image2.cpp | 4 +- tools/python/src/image4.cpp | 6 +- 14 files changed, 351 insertions(+), 44 deletions(-)
Add tfidf_transformer (#3178)
dlib/statistics.h | 1 + dlib/statistics/tfidf.h | 264 ++++++++++++++++++++++++++++++++++++++ dlib/statistics/tfidf_abstract.h | 230 +++++++++++++++++++++++++++++++++ dlib/svm/sparse_vector.h | 33 +++++ dlib/svm/sparse_vector_abstract.h | 31 +++++ dlib/test/CMakeLists.txt | 1 + dlib/test/makefile | 1 + dlib/test/sparse_vector.cpp | 36 ++++++ dlib/test/tfidf.cpp | 242 ++++++++++++++++++++++++++++++++++ 9 files changed, 839 insertions(+)
Speed up global optimization candidate selection with early rejection
.../global_optimization/global_function_search.cpp | 88 ++++++++++++++++++++- dlib/global_optimization/upper_bound_function.h | 7 ++ dlib/test/global_optimization.cpp | 89 ++++++++++++++++++++++ 3 files changed, 180 insertions(+), 4 deletions(-)
Merge pull request #3170 from lippinj/upstream-fix-2
Merge pull request #3171 from lippinj/upstream-fix-3
Merge pull request #3175 from lippinj/upstream-fix-6
Merge pull request #3169 from lippinj/upstream-fix-1
Merge pull request #3174 from lippinj/upstream-fix-5
Merge pull request #3173 from lippinj/upstream-fix-7
Merge pull request #3172 from lippinj/upstream-fix-4
Preserve Gaussian filter cutoff without dividing by tails
dlib/image_transforms/spatial_filtering.h | 4 +- dlib/test/image.cpp | 86 +++++++++++++++++++++++++++++++ 2 files changed, 87 insertions(+), 3 deletions(-)
Fix correlation normalization without epsilon regularization
dlib/statistics/statistics.h | 14 ++-- dlib/statistics/statistics_abstract.h | 2 + dlib/test/statistics.cpp | 140 ++++++++++++++++++++++++++++++++++ 3 files changed, 151 insertions(+), 5 deletions(-)
Initialize kkmeans assignments with an invalid cluster
dlib/svm/kkmeans.h | 5 ++- dlib/test/svm.cpp | 98 ++++++++++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 102 insertions(+), 1 deletion(-)
Test max-point interpolation and update tracker reference score
dlib/test/correlation_tracker.cpp | 2 +- dlib/test/matrix.cpp | 63 +++++++++++++++++++++++++++++++++++++++ 2 files changed, 64 insertions(+), 1 deletion(-)
fix(optimization): prevent integer overflow when computing oca stop time
dlib/optimization/optimization_oca.h | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
fix: remove unused variables that produce compiler warnings
dlib/geometry/point_transforms.h | 4 +--- dlib/image_processing/scan_image_pyramid.h | 2 -- 2 files changed, 1 insertion(+), 5 deletions(-)
fix(simd): make simd4f::sum() consistent across all SIMD paths
dlib/simd/simd4f.h | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-)
fix(image_transforms): guard against division by zero in create_gaussian_filter()
dlib/image_transforms/spatial_filtering.h | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)
fix(svm): initialize assignments vector after resize in kkmeans
dlib/svm/kkmeans.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
fix(statistics): guard against division by zero in correlation()
dlib/statistics/statistics.h | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)
fix(matrix): correct column index in max-point interpolation
dlib/matrix/matrix_la.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
fix bug in unit test
dlib/test/dnn.cpp | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
cblas: update cmake_minimum_required to 3.10.0
dlib/external/cblas/CMakeLists.txt | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Add support for building dlib on Windows ARM64 (#3164)
.../set_compiler_specific_options.cmake | 18 ++- dlib/cmake_utils/test_for_simd_options.cmake | 37 ++++++ dlib/simd/simd4f.h | 17 ++- dlib/simd/simd4i.h | 28 ++++- dlib/simd/simd_check.h | 6 +- dlib/test/CMakeLists.txt | 1 + dlib/test/simd.cpp | 129 +++++++++++++++++++++ tools/python/test/test_setup.py | 46 ++++++++ 8 files changed, 274 insertions(+), 8 deletions(-)
Let's start using C++17 (#3167)
dlib/CMakeLists.txt | 7 +++++-- dlib/algs.h | 16 +++++++--------- dlib/java/cmake_swig_jni | 2 +- dlib/serialize.h | 2 +- dlib/test/WINDOWS_build_and_run_all_unit_tests.bat | 12 ++++++------ dlib/test/constexpr_if.cpp | 2 +- dlib/test/makefile | 2 +- 7 files changed, 22 insertions(+), 21 deletions(-)
Fall back to the cuDNN heuristics when benchmarking runs out of memory (#3166)
dlib/cuda/cudnn_dlibapi.cpp | 87 ++++++++++++++++++++++++++++++++++++++------- 1 file changed, 75 insertions(+), 12 deletions(-)
Choose each convolution algorithm when first needed (#3162)
dlib/cuda/cudnn_dlibapi.cpp | 331 +++++++++++++++++++++++++------------------- dlib/cuda/cudnn_dlibapi.h | 15 +- 2 files changed, 197 insertions(+), 149 deletions(-)
Optionally choose the convolution algorithms for each input shape (#3161)
dlib/cuda/cudnn_dlibapi.cpp | 67 +++++++++++++++++++++++++++++++++----- dlib/cuda/tensor_tools.cpp | 20 ++++++++++++ dlib/cuda/tensor_tools.h | 2 ++ dlib/dnn/core_abstract.h | 25 +++++++++++++++ dlib/test/dnn.cpp | 78 +++++++++++++++++++++++++++++++++++++++++++++ 5 files changed, 184 insertions(+), 8 deletions(-)
Use the C++ tests' tolerance in the Holder table Python test (#3163)
tools/python/test/test_global_optimization.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Add CUDA 13 config removing SM_70 (#3160)
dlib/CMakeLists.txt | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-)
update libpng and zlib
dlib/external/README.png.md | 30 + dlib/external/libpng/LICENSE | 199 +- dlib/external/libpng/README | 309 +- dlib/external/libpng/arm/arm_init.c | 166 +- dlib/external/libpng/arm/filter_neon.S | 267 +- dlib/external/libpng/arm/filter_neon_intrinsics.c | 67 +- dlib/external/libpng/arm/palette_neon_intrinsics.c | 78 +- dlib/external/libpng/png.c | 1539 +-- dlib/external/libpng/png.h | 1269 ++- dlib/external/libpng/pngconf.h | 68 +- dlib/external/libpng/pngdebug.h | 12 +- dlib/external/libpng/pngerror.c | 205 +- dlib/external/libpng/pngget.c | 590 +- dlib/external/libpng/pnginfo.h | 104 +- dlib/external/libpng/pnglibconf.h | 25 +- dlib/external/libpng/pngmem.c | 59 +- dlib/external/libpng/pngpread.c | 213 +- dlib/external/libpng/pngpriv.h | 1512 +-- dlib/external/libpng/pngread.c | 700 +- dlib/external/libpng/pngrio.c | 5 +- dlib/external/libpng/pngrtran.c | 541 +- dlib/external/libpng/pngrutil.c | 1906 ++-- dlib/external/libpng/pngset.c | 493 +- dlib/external/libpng/pngstruct.h | 123 +- dlib/external/libpng/pngtrans.c | 57 +- dlib/external/libpng/pngwio.c | 9 +- dlib/external/libpng/pngwrite.c | 235 +- dlib/external/libpng/pngwtran.c | 1 - dlib/external/libpng/pngwutil.c | 203 +- dlib/external/zlib/README | 44 +- dlib/external/zlib/adler32.c | 32 +- dlib/external/zlib/compress.c | 61 +- dlib/external/zlib/crc32.c | 1193 ++- dlib/external/zlib/crc32.h | 9877 +++++++++++++++++++- dlib/external/zlib/deflate.c | 1028 +- dlib/external/zlib/deflate.h | 80 +- dlib/external/zlib/gzclose.c | 4 +- dlib/external/zlib/gzguts.h | 102 +- dlib/external/zlib/gzlib.c | 216 +- dlib/external/zlib/gzread.c | 396 +- dlib/external/zlib/gzwrite.c | 369 +- dlib/external/zlib/infback.c | 133 +- dlib/external/zlib/inffast.c | 46 +- dlib/external/zlib/inffast.h | 2 +- dlib/external/zlib/inffixed.h | 182 +- dlib/external/zlib/inflate.c | 367 +- dlib/external/zlib/inflate.h | 7 +- dlib/external/zlib/inftrees.c | 154 +- dlib/external/zlib/inftrees.h | 16 +- dlib/external/zlib/trees.c | 692 +- dlib/external/zlib/uncompr.c | 74 +- dlib/external/zlib/zconf.h | 67 +- dlib/external/zlib/zlib.h | 845 +- dlib/external/zlib/zutil.c | 144 +- dlib/external/zlib/zutil.h | 150 +- 55 files changed, 18508 insertions(+), 8758 deletions(-)
make this test more robust
dlib/test/timer.cpp | 271 ++++++++++++++++++++++++++++++++++++++-------------- 1 file changed, 199 insertions(+), 72 deletions(-)
Gate PUBLIC compiler-specific flags by consumer compiler ID (#3157)
dlib/CMakeLists.txt | 13 ++-- dlib/cmake_utils/dlibConfig.cmake.in | 72 ++++++++++++++++++++-- .../set_compiler_specific_options.cmake | 15 +++-- .../test_for_dlib_config/CMakeLists.txt | 37 +++++++++++ .../test_for_dlib_config/dlib_config_test.cpp | 10 +++ 5 files changed, 132 insertions(+), 15 deletions(-)
Pybind11 free threading (#3158)
dlib/external/pybind11/CMakeLists.txt | 176 +- dlib/external/pybind11/README.dlib | 13 + dlib/external/pybind11/README.rst | 84 +- dlib/external/pybind11/include/pybind11/attr.h | 62 +- .../pybind11/include/pybind11/buffer_info.h | 35 +- dlib/external/pybind11/include/pybind11/cast.h | 867 ++++++++-- dlib/external/pybind11/include/pybind11/chrono.h | 4 +- dlib/external/pybind11/include/pybind11/complex.h | 23 +- .../pybind11/include/pybind11/conduit/README.txt | 15 + .../include/pybind11/conduit/pybind11_conduit_v1.h | 116 ++ .../pybind11/conduit/pybind11_platform_abi_id.h | 87 + .../pybind11/conduit/wrap_include_python_h.h | 72 + .../pybind11/include/pybind11/critical_section.h | 56 + .../include/pybind11/detail/argument_vector.h | 417 +++++ .../pybind11/include/pybind11/detail/class.h | 322 ++-- .../pybind11/include/pybind11/detail/common.h | 558 +++--- .../pybind11/include/pybind11/detail/cpp_conduit.h | 75 + .../pybind11/include/pybind11/detail/descr.h | 55 + .../detail/dynamic_raw_ptr_cast_if_possible.h | 39 + .../pybind11/detail/exception_translation.h | 71 + .../pybind11/detail/function_record_pyobject.h | 192 +++ .../detail/holder_caster_foreign_helpers.h | 104 ++ .../pybind11/include/pybind11/detail/init.h | 170 +- .../pybind11/include/pybind11/detail/internals.h | 1121 ++++++++---- .../include/pybind11/detail/native_enum_data.h | 227 +++ .../pybind11/detail/pybind11_namespace_macros.h | 82 + .../include/pybind11/detail/struct_smart_holder.h | 398 +++++ .../include/pybind11/detail/type_caster_base.h | 917 +++++++--- .../include/pybind11/detail/using_smart_holder.h | 22 + .../include/pybind11/detail/value_and_holder.h | 92 + .../pybind11/include/pybind11/eigen/matrix.h | 29 +- .../pybind11/include/pybind11/eigen/tensor.h | 23 +- dlib/external/pybind11/include/pybind11/embed.h | 96 +- dlib/external/pybind11/include/pybind11/eval.h | 17 +- .../pybind11/include/pybind11/functional.h | 107 +- dlib/external/pybind11/include/pybind11/gil.h | 120 +- .../pybind11/include/pybind11/gil_safe_call_once.h | 192 ++- .../pybind11/include/pybind11/gil_simple.h | 37 + dlib/external/pybind11/include/pybind11/iostream.h | 44 +- .../pybind11/include/pybind11/native_enum.h | 76 + dlib/external/pybind11/include/pybind11/numpy.h | 467 +++-- dlib/external/pybind11/include/pybind11/pybind11.h | 1775 +++++++++++++++----- dlib/external/pybind11/include/pybind11/pytypes.h | 250 ++- dlib/external/pybind11/include/pybind11/stl.h | 336 +++- .../pybind11/include/pybind11/stl/filesystem.h | 48 +- dlib/external/pybind11/include/pybind11/stl_bind.h | 51 +- .../pybind11/include/pybind11/subinterpreter.h | 291 ++++ .../pybind11/trampoline_self_life_support.h | 65 + dlib/external/pybind11/include/pybind11/typing.h | 186 +- dlib/external/pybind11/include/pybind11/warnings.h | 75 + dlib/external/pybind11/tools/FindCatch.cmake | 2 - .../pybind11/tools/FindPythonLibsNew.cmake | 14 +- .../tools/codespell_ignore_lines_from_errors.py | 5 +- dlib/external/pybind11/tools/libsize.py | 2 + dlib/external/pybind11/tools/make_changelog.py | 95 +- dlib/external/pybind11/tools/make_global.py | 33 + dlib/external/pybind11/tools/pybind11Common.cmake | 155 +- .../pybind11/tools/pybind11Config.cmake.in | 13 +- .../tools/pybind11GuessPythonExtSuffix.cmake | 94 ++ .../external/pybind11/tools/pybind11NewTools.cmake | 168 +- dlib/external/pybind11/tools/pybind11Tools.cmake | 56 +- dlib/external/pybind11/tools/pyproject.toml | 3 - dlib/external/pybind11/tools/setup_global.py.in | 63 - dlib/external/pybind11/tools/setup_main.py.in | 44 - .../tools/test-pybind11GuessPythonExtSuffix.cmake | 185 ++ tools/python/CMakeLists.txt | 14 +- tools/python/src/cnn_face_detector.cpp | 5 + tools/python/src/dlib.cpp | 2 +- tools/python/src/face_recognition.cpp | 6 +- tools/python/src/numpy_returns.cpp | 2 +- tools/python/test/test_numpy_returns.py | 11 + 71 files changed, 9419 insertions(+), 2310 deletions(-)
Add dependency-free ONNX export for DNNs (#3145)
dlib/dnn.h | 3 +- dlib/dnn/onnx.h | 2089 ++++++++++++++++++++++++++++++ dlib/dnn/onnx_abstract.h | 127 ++ dlib/test/dnn.cpp | 511 +++++++- dlib/test/onnx_test_helpers.h | 494 +++++++ examples/CMakeLists.txt | 1 + examples/dnn_imagenet_export_onnx_ex.cpp | 118 ++ 7 files changed, 3340 insertions(+), 3 deletions(-)
Allow setting sockstreambuf buffer size and max putback (#3154)
dlib/sockstreambuf/sockstreambuf.cpp | 10 ++-- dlib/sockstreambuf/sockstreambuf.h | 74 ++++++++++++----------- dlib/sockstreambuf/sockstreambuf_abstract.h | 28 ++++++++- dlib/test/sockstreambuf.cpp | 93 +++++++++++++++++++++++++++++ 4 files changed, 164 insertions(+), 41 deletions(-)
Added support for unix sockets (#3152)
dlib/sockets/sockets_extensions.cpp | 18 + dlib/sockets/sockets_extensions.h | 6 + dlib/sockets/sockets_extensions_abstract.h | 16 + dlib/sockets/sockets_kernel_1.cpp | 517 ++++++++++++++++++++++------ dlib/sockets/sockets_kernel_1.h | 160 ++++++++- dlib/sockets/sockets_kernel_2.cpp | 527 +++++++++++++++++++++-------- dlib/sockets/sockets_kernel_2.h | 137 +++++++- dlib/sockets/sockets_kernel_abstract.h | 192 ++++++++++- dlib/test/sockets2.cpp | 86 ++++- 9 files changed, 1388 insertions(+), 271 deletions(-)
use auto for sorted_rects iterator (#3150)
dlib/image_processing/scan_image_pyramid_tools.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
lbfgs_search_strategy: Fix gcc 16 warning (#3144)
dlib/optimization/optimization_search_strategies.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Support for runtime CPU/CUDA selection (#3060)
dlib/cuda/cuda_dlib.cu | 69 ++- dlib/cuda/cuda_dlib.h | 6 + dlib/cuda/curand_dlibapi.cpp | 12 +- dlib/cuda/gpu_data.cpp | 22 +- dlib/cuda/gpu_data.h | 24 +- dlib/cuda/gpu_data_abstract.h | 15 +- dlib/cuda/tensor.h | 57 +- dlib/cuda/tensor_abstract.h | 12 +- dlib/cuda/tensor_tools.cpp | 1183 +++++++++++++++++++++++------------------ dlib/cuda/tensor_tools.h | 170 +++++- dlib/dnn/core_abstract.h | 17 + dlib/dnn/loss.h | 60 ++- dlib/test/cublas.cpp | 2 + dlib/test/dnn.cpp | 911 +++++++++++++++---------------- 14 files changed, 1508 insertions(+), 1052 deletions(-)
Do some cleanup
dlib/data_io/image_dataset_metadata.cpp | 32 ++++++++++++++- dlib/data_io/image_dataset_metadata.h | 37 ++++++++--------- dlib/test/data_io.cpp | 72 ++++++++++++++++++++++++++++++++- tools/imglab/src/metadata_editor.cpp | 48 ++++++++++++++++------ tools/imglab/src/metadata_editor.h | 11 +++-- 5 files changed, 159 insertions(+), 41 deletions(-)
Add undo and redo functionality to imglab (#606) (#3143)
dlib/data_io/image_dataset_metadata.h | 21 ++++++++++ tools/imglab/src/metadata_editor.cpp | 79 +++++++++++++++++++++++++++++++++++ tools/imglab/src/metadata_editor.h | 7 ++++ 3 files changed, 107 insertions(+)
set back to .99
dlib/CMakeLists.txt | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
tag 20.0.1
dlib/CMakeLists.txt | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
fix(test/string): fix gcc 16 build issue (#3137)
dlib/test/string.cpp | 4 ++-- dlib/test/tokenizer.cpp | 8 ++++---- 2 files changed, 6 insertions(+), 6 deletions(-)
Improve numerical robustness of find_min_trust_region()
dlib/optimization/optimization_trust_region.h | 10 +++- dlib/test/trust_region.cpp | 75 +++++++++++++++++++++++++++ 2 files changed, 83 insertions(+), 2 deletions(-)
Update path to mkl and kiss fft headers (#3136)
dlib/test/fft.cpp | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
Problem: on Blackwell GPUs, functions such as `cudnnFindConvolutionBackwardDataAlgorithm` may run out of stack space on Windows (leading to a stack overflow error) (#3130)
examples/CMakeLists.txt | 10 ++++++++++ 1 file changed, 10 insertions(+)
Fix cuDNN convolution precision on Ampere+ GPUs (#3127)
dlib/cuda/cudnn_dlibapi.cpp | 9 +++++++++ 1 file changed, 9 insertions(+)
Set test threshold more appropriately so it doesn't randomly fail.
dlib/test/dnn.cpp | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-)
Fix flaky holder table test by increasing max function calls (#3129)
tools/python/test/test_global_optimization.py | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
Fix ACT layer gradient computation on CUDA (#3128)
dlib/cuda/cpu_dlib.cpp | 7 +++++++ dlib/cuda/cpu_dlib.h | 2 ++ dlib/cuda/cuda_dlib.cu | 9 +++++++++ dlib/cuda/cuda_dlib.h | 2 ++ dlib/cuda/tensor_tools.cpp | 10 ++++++---- dlib/cuda/tensor_tools.h | 9 ++++++--- dlib/dnn/layers.h | 30 ++---------------------------- 7 files changed, 34 insertions(+), 35 deletions(-)
Fix callable detection to support functools.partial and all callable objects (#3121)
tools/python/src/global_optimization.cpp | 44 ++++++++++++++++++++++++++++---- 1 file changed, 39 insertions(+), 5 deletions(-)
Implementation of ARC-AGI dataset loader (#3115)
dlib/data_io.h | 1 + dlib/data_io/arc_agi.h | 909 ++++++++++++++++++++++++++++++++++++++++ dlib/data_io/arc_agi_abstract.h | 302 +++++++++++++ 3 files changed, 1212 insertions(+)
Improve numerical stability of batch normalization on CPUs (#3113)
dlib/cuda/cpu_dlib.cpp | 90 ++++++++++++++++++++++++++------------------------ dlib/test/dnn.cpp | 4 +-- 2 files changed, 48 insertions(+), 46 deletions(-)
Add Adaptive Computation Time (ACT) layer (#3111)
dlib/cuda/cpu_dlib.cpp | 147 +++++++++++++++++ dlib/cuda/cpu_dlib.h | 48 ++++++ dlib/cuda/cuda_dlib.cu | 260 ++++++++++++++++++++++++++++- dlib/cuda/cuda_dlib.h | 48 ++++++ dlib/cuda/tensor_tools.cpp | 84 ++++++++++ dlib/cuda/tensor_tools.h | 126 +++++++++++++- dlib/dnn/layers.h | 398 ++++++++++++++++++++++++++++++++++++++++++++- dlib/dnn/layers_abstract.h | 119 ++++++++++++++ dlib/test/dnn.cpp | 203 ++++++++++++++--------- 9 files changed, 1357 insertions(+), 76 deletions(-)
Improve the somewhat flaky `test_loss_multibinary_log` by avoiding samples very close to class boundaries (#3112)
dlib/test/dnn.cpp | 45 +++++++++++++++++++++++++++++++++++++++++---- 1 file changed, 41 insertions(+), 4 deletions(-)
Replace FindCUDA with FindCUDAToolkit (fix #2833) (#3090)
dlib/CMakeLists.txt | 249 +++++++------------------ dlib/cmake_utils/FindCUDNN.cmake | 81 ++++++++ dlib/cmake_utils/dlibConfig.cmake.in | 4 + dlib/cmake_utils/test_for_cuda/CMakeLists.txt | 14 -- dlib/cmake_utils/test_for_cuda/cuda_test.cu | 21 --- dlib/cmake_utils/test_for_cudnn/CMakeLists.txt | 18 -- dlib/cmake_utils/test_for_cudnn/find_cudnn.txt | 24 --- tools/python/dlib/__init__.py.in | 15 +- 8 files changed, 161 insertions(+), 265 deletions(-)
[unicode] fixed warning in convert_mbstring_to_wstring() (#3110)
dlib/unicode/unicode.cpp | 33 +++++++++++++++++++++++++-------- 1 file changed, 25 insertions(+), 8 deletions(-)
FFMpeg 7.1 API (#3105)
dlib/media/ffmpeg_muxer.h | 92 +++++++++++++++++++++++++++++++++-------------- 1 file changed, 65 insertions(+), 27 deletions(-)
fix -Wcast-user-defined (#3103)
dlib/dnn/layers.h | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-)
update python packaging to work with newer toolchains (#3102)
tools/python/CMakeLists.txt | 7 ------- 1 file changed, 7 deletions(-)
Suppress a spurious warning in unit tests and make colors not always output
dlib/test/CMakeLists.txt | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)
fix up build errors with the regular makefile and non-cmake builds.
dlib/all/source.cpp | 4 +++- dlib/image_loader/png_loader.h | 10 ++-------- dlib/test/makefile | 2 +- 3 files changed, 6 insertions(+), 10 deletions(-)
std::aligned_storage_t is deprecated in c++23 (#3100)
dlib/any/storage.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
Account for change in ref count in Python 3.14+ (#3098)
tools/python/test/test_numpy_returns.py | 17 +++++++++++++++-- 1 file changed, 15 insertions(+), 2 deletions(-)
cleanup
dlib/image_processing/full_object_detection.h | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-)
full_object_detection: API Support for dpoint (#2183) (#3095)
dlib/data_io/load_image_dataset.h | 2 +- dlib/image_processing/full_object_detection.h | 40 ++++++++++++++++------ .../full_object_detection_abstract.h | 22 ++++++++---- dlib/image_processing/scan_image_pyramid.h | 2 +- dlib/image_transforms/interpolation.h | 2 +- dlib/test/object_detector.cpp | 2 +- 6 files changed, 49 insertions(+), 21 deletions(-)
updated docs
dlib/dnn/layers_abstract.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Add transformer example with RoPE and MoE-like mechanisms (#3078)
dlib/dnn/layers.h | 37 +- dlib/tokenizer/bpe_tokenizer.h | 697 +++++++++------- dlib/tokenizer/bpe_tokenizer_abstract.h | 46 +- examples/CMakeLists.txt | 1 + examples/slm_advanced_train_ex.cpp | 1362 +++++++++++++++++++++++++++++++ 5 files changed, 1812 insertions(+), 331 deletions(-)
`throw()` -> `noexcept` (#3092)
dlib/any/any_abstract.h | 2 +- dlib/cmd_line_parser/cmd_line_parser_check_1.h | 2 +- dlib/cmd_line_parser/cmd_line_parser_kernel_1.h | 2 +- dlib/config_reader/config_reader_kernel_1.h | 4 ++-- dlib/error.h | 6 +++--- dlib/misc_api/misc_api_kernel_1.h | 2 +- dlib/std_allocator.h | 14 +++++++------- dlib/svm/one_vs_all_trainer.h | 2 +- dlib/svm/one_vs_one_trainer.h | 2 +- dlib/type_safe_union/type_safe_union_kernel.h | 2 +- 10 files changed, 19 insertions(+), 19 deletions(-)
Problem: reading large images (in my case, it was a 40792x20480 RGBA PNG one) causes a crash, when the product row-number * width-step overflows a 32-bit integer (#3091)
dlib/array2d/array2d_generic_image.h | 2 +- dlib/array2d/array2d_kernel.h | 2 +- dlib/array2d/array2d_kernel_abstract.h | 2 +- dlib/image_processing/generic_image.h | 16 ++++++++-------- dlib/image_transforms/interpolation.h | 8 ++++---- dlib/matrix/matrix_generic_image.h | 2 +- dlib/opencv/cv_image.h | 4 ++-- dlib/opencv/cv_image_abstract.h | 2 +- dlib/python/numpy_image.h | 2 +- 9 files changed, 20 insertions(+), 20 deletions(-)
fix the miss of including matlab ut library (#3089)
dlib/matlab/cmake_mex_wrapper | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
Fix: Correct YOLO loss gradient for x,y coordinates (#3088)
dlib/dnn/loss.h | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
Minor optimization: `emplace_back` neighbors (#3086)
dlib/image_transforms/label_connected_blobs.h | 26 +++++++++++++------------- 1 file changed, 13 insertions(+), 13 deletions(-)
remove std::move where clang warns of pessimizing-move (#3080)
dlib/any/any.h | 2 +- dlib/any/storage.h | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
Record last changeset and set PATCH version to 99
dlib/CMakeLists.txt | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Created release v20.0
dlib/CMakeLists.txt | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)