dlib C++ Library - Change Log

Change Log

Revision: 70ea028f12e10f4d992a
Author: Davis King
Date: Mar 28, 2021 (05:17:49 UTC)

Created release v19.22

 dlib/CMakeLists.txt | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

Revision: a44ddd7452221733dec9
Author: Adrià Arrufat
Date: Mar 22, 2021 (29:37:54 UTC)

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(+)

Revision: 092fa3037f1004ead614
Author: Adrià Arrufat
Date: Mar 08, 2021 (21:59:53 UTC)

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(+)

Revision: 0697527acc9490f3141c
Author: Davis King
Date: Feb 16, 2021 (15:31:55 UTC)

fix invalid assert

 dlib/global_optimization/global_function_search.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Revision: 1b58bdc205234e753404
Author: pfeatherstone
Date: Feb 13, 2021 (04:42:47 UTC)

[SERIALIZATION] updated _abstract file (#2306)

 dlib/vectorstream/vectorstream_abstract.h | 22 ++++++++++++++--------
 1 file changed, 14 insertions(+), 8 deletions(-)

Revision: 479b69e68863def21ed0
Author: pfeatherstone
Date: Feb 12, 2021 (03:13:05 UTC)

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(-)

Revision: 04a3534af1690b88f4aa
Author: Adrià Arrufat
Date: Feb 11, 2021 (20:55:54 UTC)

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(-)

Revision: 42e6ace845fbccba5bd7
Author: Adrià Arrufat
Date: Feb 09, 2021 (31:07:24 UTC)

Minor fix in the network size format (#2303)

 dlib/dnn/trainer.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Revision: 9f6aefc0db01e1d2edca
Author: Davis King
Date: Jan 28, 2021 (03:35:17 UTC)

Add missing .get_final_data_gradient() for repeat layer

 dlib/dnn/core.h | 3 +++
 1 file changed, 3 insertions(+)

Revision: 1516cf31c3ab05e6c62b
Author: pfeatherstone
Date: Jan 18, 2021 (13:58:30 UTC)

[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(-)

Revision: 22007f6665f5ea99ab63
Author: Davis King
Date: Jan 18, 2021 (03:56:34 UTC)

work around bug in visual studio

 dlib/test/global_optimization.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Revision: b8807922ddbe66ad7b06
Author: pfeatherstone
Date: Jan 17, 2021 (18:22:06 UTC)

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(-)

Revision: 044ff91b797e40e0b31c
Author: Davis King
Date: Jan 17, 2021 (08:18:15 UTC)

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(-)

Revision: 95271cfe43ffceeadeb1
Author: Robert Underwood
Date: Jan 17, 2021 (07:59:57 UTC)

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(-)

Revision: 600e03655274f0e7f201
Author: Davis King
Date: Jan 16, 2021 (05:13:52 UTC)

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(-)

Revision: df46c516b9e493e0e6fa
Author: Davis King
Date: Jan 16, 2021 (-05:42:29 UTC)

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(-)

Revision: 4c0b8b96906d3a1a1a14
Author: Davis King
Date: Jan 16, 2021 (-05:41:30 UTC)

Fix compiler warning

 dlib/cuda/cuda_dlib.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Revision: 0871081a231f06ac8652
Author: Robert Underwood
Date: Jan 14, 2021 (17:05:15 UTC)

fix find_min_global finding wrong answers (#2284)

 dlib/global_optimization/find_max_global.h | 5 +++++
 1 file changed, 5 insertions(+)

Revision: 4a53742f8f6a071fb002
Author: pkubaj
Date: Jan 13, 2021 (00:55:28 UTC)

Fix build with clang on ppc64le (#2278)

 dlib/simd/simd4f.h | 5 +++++
 1 file changed, 5 insertions(+)

Revision: 9e5c556740dc84616027
Author: Davis King
Date: Jan 06, 2021 (02:48:16 UTC)

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(-)

Revision: 39eec294b01c96c708f7
Author: pfeatherstone
Date: Jan 06, 2021 (03:19:05 UTC)

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(-)

Revision: 71dd9a6cd7f2512a82ff
Author: Adrià Arrufat
Date: Jan 03, 2021 (31:29:09 UTC)

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(-)

Revision: 187b73f7dc0566c38665
Author: Adrià Arrufat
Date: Dec 31, 2020 (11:28:53 UTC)

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(-)

Revision: 74baff132a68d2388a7d
Author: stoperro
Date: Dec 28, 2020 (17:37:17 UTC)

Fixed CUDA 11.1 compilation. (#2271)

 dlib/cuda/cuda_dlib.cu | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)

Revision: 20378b0819e37ff8694e
Author: Davis King
Date: Dec 18, 2020 (15:12:12 UTC)

make test run faster

 dlib/test/rand.cpp | 25 +++++++++++--------------
 1 file changed, 11 insertions(+), 14 deletions(-)

Revision: 0fede57ed7ef756e9812
Author: Davis King
Date: Dec 16, 2020 (15:37:22 UTC)

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(-)

Revision: c5c761b5ad29ff7692d6
Author: Davis King
Date: Dec 13, 2020 (15:36:06 UTC)

Fix build error from ambiguous calls to call_if_valid()

 dlib/metaprogramming.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

Revision: 286ca71e1a521ef5f24b
Author: Akshay Naresh Modi
Date: Dec 13, 2020 (03:57:41 UTC)

Release GIL before detecting faces (#2252)

 tools/python/src/simple_object_detector_py.h | 18 ++++++++++++++----
 1 file changed, 14 insertions(+), 4 deletions(-)

Revision: 7b5b3750260347591746
Author: Abdolkarim Saeedi
Date: Dec 09, 2020 (19:37:45 UTC)

Update dnn_inception_ex.cpp (#2256)

 examples/dnn_inception_ex.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Revision: eb1f08ce6ab3ca6f9d10
Author: yuriio
Date: Dec 07, 2020 (04:05:07 UTC)

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(-)

Revision: f42a6d23cc67c985fc9b
Author: Davis King
Date: Dec 04, 2020 (09:18:21 UTC)

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(+)

Revision: 0ff61299b77236c9125b
Author: pfeatherstone
Date: Dec 04, 2020 (03:05:39 UTC)

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(+)

Revision: e58da5135c15f0dcc404
Author: Davis King
Date: Nov 29, 2020 (09:53:08 UTC)

minor cleanup

 dlib/data_io/cifar.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Revision: d3b0213118769069c1e7
Author: Adrià Arrufat
Date: Nov 30, 2020 (13:47:34 UTC)

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(-)

Revision: d9e58d66cfe15f8189dd
Author: pfeatherstone
Date: Nov 25, 2020 (03:09:38 UTC)

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(-)

Revision: a7627cbd07cb196fb9f8
Author: Adrià Arrufat
Date: Nov 25, 2020 (21:07:04 UTC)

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(-)

Revision: b6bf8aefee5692c8da0a
Author: Adrià Arrufat
Date: Nov 22, 2020 (16:09:06 UTC)

Add support for matrix serialization to python API (#2241)

 tools/python/src/matrix.cpp | 12 ++++++++++++
 1 file changed, 12 insertions(+)

Revision: d7644ef2b7234520ddcf
Author: Frankie Robertson
Date: Nov 22, 2020 (02:07:28 UTC)

Expose get_face_chip_details to Python (#2238)

 tools/python/src/image2.cpp | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)

Revision: 96a75568be7701dedb94
Author: Adrià Arrufat
Date: Nov 18, 2020 (31:26:12 UTC)

fix unused parameter warning in visitor_net_to_xml (#2240)

 dlib/dnn/utilities.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

Revision: 820fd353d25b0c8cb206
Author: Adrià Arrufat
Date: Nov 17, 2020 (21:17:29 UTC)

Make dnn_trainer print the minibatch size to ostream (#2236)

 dlib/dnn/trainer.h | 1 +
 1 file changed, 1 insertion(+)

Revision: 375f117222e13152c00f
Author: Adrià Arrufat
Date: Nov 15, 2020 (30:55:54 UTC)

Add custom ostream to console progress indicator (#2234)

 dlib/console_progress_indicator.h | 36 +++++++++++++++++++-----------------
 1 file changed, 19 insertions(+), 17 deletions(-)

Revision: 2ef8e3ac14671cf8a356
Author: Adrià Arrufat
Date: Nov 13, 2020 (21:39:15 UTC)

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(-)

Revision: 93b992d79024b68544cf
Author: Davis King
Date: Nov 08, 2020 (05:25:01 UTC)

slightly improve tests

 dlib/test/pixel.cpp | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

Revision: 3f163bd4332351d8e93c
Author: Davis King
Date: Nov 08, 2020 (05:22:40 UTC)

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(-)

Revision: 83921b390e6cf7575b99
Author: Adrià Arrufat
Date: Oct 21, 2020 (31:15:08 UTC)

Remove an unused variable and old commented code (#2217)

 dlib/cuda/cuda_dlib.cu | 3 ---
 1 file changed, 3 deletions(-)

Revision: 3c82c2259c558d2ae9bb
Author: Adrià Arrufat
Date: Oct 20, 2020 (29:56:55 UTC)

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(-)

Revision: 5074850356e76a4c601c
Author: Adrià Arrufat
Date: Oct 14, 2020 (30:17:30 UTC)

fix backtracking when losses stay at inf (fixes #2206) (#2209)

 dlib/dnn/trainer.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

Revision: a1f158379e2f328e8697
Author: Adrià Arrufat
Date: Oct 10, 2020 (30:42:10 UTC)

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(-)

Revision: 3ba004f875d02c4c326f
Author: Adrià Arrufat
Date: Oct 09, 2020 (20:45:23 UTC)

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(+)

Revision: f4f8bff95ec84d441953
Author: Davis King
Date: Oct 08, 2020 (18:33:35 UTC)

fix cmake not finding openblas

 dlib/cmake_utils/find_blas.cmake | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

Revision: c45d166a2592059d573e
Author: Adrià Arrufat
Date: Oct 06, 2020 (19:20:37 UTC)

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(-)

Revision: d78d273a459107f12424
Author: Adrià Arrufat
Date: Sep 30, 2020 (30:04:28 UTC)

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(-)

Revision: 4125a7bb1f07ceff1b91
Author: pfeatherstone
Date: Sep 25, 2020 (14:27:30 UTC)

DLIB (de)serialization : enhanced STL container support (#2185)

 dlib/serialize.h        | 660 +++++++++++++++++++++++++++++++++++++++++++++++-
 dlib/test/serialize.cpp |  91 ++++++-
 2 files changed, 737 insertions(+), 14 deletions(-)

Revision: 5408b17f74081520f1f4
Author: aviezab
Date: Sep 25, 2020 (25:48:48 UTC)

Linux Distro Detection to fix issue number #2159 #154 (#2169)

 dlib/cmake_utils/find_blas.cmake | 13 +++++++------
 1 file changed, 7 insertions(+), 6 deletions(-)

Revision: 0419b816893f314b67db
Author: Davis King
Date: Sep 25, 2020 (03:41:26 UTC)

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(-)

Revision: 20a1477209e1fd28b0f1
Author: Davis King
Date: Sep 19, 2020 (03:21:52 UTC)

update docs

 dlib/serialize.h | 10 ++++++++++
 1 file changed, 10 insertions(+)

Revision: ab346ddfa640bb0a8b75
Author: pfeatherstone
Date: Sep 19, 2020 (13:16:21 UTC)

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(-)

Revision: fa818b9a964cba9a093d
Author: Adrià Arrufat
Date: Sep 18, 2020 (20:54:06 UTC)

use DLIB_CASSERT to avoid unused variable warning in release compilation (#2182)

 dlib/vectorstream/vectorstream.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

Revision: d4fe74b5a82ef21e7216
Author: pfeatherstone
Date: Sep 17, 2020 (02:37:36 UTC)

vectorstream updates: added seekoff and seekpos (#2179)

 dlib/test/vectorstream.cpp       | 81 ++++++++++++++++++++++++++++++++++------
 dlib/vectorstream/vectorstream.h | 41 +++++++++++++++-----
 2 files changed, 101 insertions(+), 21 deletions(-)

Revision: cdeb2e067cf99ce52b60
Author: Davis King
Date: Sep 12, 2020 (17:52:21 UTC)

add some docs

 dlib/serialize.h | 102 +++++++++++++++++++++++++++++++++++++++----------------
 1 file changed, 73 insertions(+), 29 deletions(-)

Revision: 12a82f6542ee0fcb66f5
Author: pfeatherstone
Date: Sep 13, 2020 (03:18:46 UTC)

Macro for generating default serialisation functions (#2177)

 dlib/serialize.h        | 72 ++++++++++++++++++++++++++++++++++++++++++++++++-
 dlib/test/serialize.cpp | 50 ++++++++++++++++++++++++++++++++++
 2 files changed, 121 insertions(+), 1 deletion(-)

Revision: 9d60949a3a937f33acfc
Author: Adrià Arrufat
Date: Sep 12, 2020 (29:55:24 UTC)

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(+)

Revision: 77e6255fdd2a6ecfbaf7
Author: Adrià Arrufat
Date: Sep 08, 2020 (29:16:15 UTC)

Add error message for mismatched tensor sizes in dnn_trainer (#2165)

 dlib/dnn/trainer.h | 5 +++++
 1 file changed, 5 insertions(+)

Revision: 40c3e48818b6decb97d3
Author: Davis King
Date: Sep 06, 2020 (06:42:56 UTC)

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(-)

Revision: 5ec60a91c4aa0fc17a73
Author: Adrià Arrufat
Date: Sep 06, 2020 (31:27:50 UTC)

Show how to use the new visitors with lambdas (#2162)

 examples/dnn_dcgan_train_ex.cpp | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

Revision: 393db2490b8e7fbe0dbe
Author: Davis King
Date: Sep 06, 2020 (04:57:44 UTC)

switch this to C++11 code

 dlib/test/metaprogramming.cpp | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

Revision: 5bcbe617eb27407bd211
Author: Davis King
Date: Sep 06, 2020 (04:45:23 UTC)

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(-)

Revision: afe19fcb8bef83335274
Author: Davis King
Date: Sep 05, 2020 (14:33:04 UTC)

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(-)

Revision: 7dcc7b4ebc65a5ae860e
Author: Davis King
Date: Sep 05, 2020 (10:38:24 UTC)

Added call_if_valid()

 dlib/metaprogramming.h        | 40 ++++++++++++++++++++++++++++++++++++++++
 dlib/test/metaprogramming.cpp | 37 +++++++++++++++++++++++++++++++++++++
 2 files changed, 77 insertions(+)

Revision: e7ec6b777757d0c589cb
Author: Adrià Arrufat
Date: Sep 03, 2020 (19:59:19 UTC)

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(-)

Revision: ed22f0400aa3b8b3a3ce
Author: Davis King
Date: Sep 02, 2020 (17:48:30 UTC)

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(-)

Revision: 0bb6ce36d8ebd16fe755
Author: Davis King
Date: Sep 02, 2020 (17:47:58 UTC)

dnn_trainer prints number of steps executed when print to ostream

 dlib/dnn/trainer.h | 1 +
 1 file changed, 1 insertion(+)

Revision: 76cc8e3b6bbb74a9b68f
Author: Davis King
Date: Sep 02, 2020 (17:42:44 UTC)

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(+)

Revision: c14ba4847e76f78a2f78
Author: Davis King
Date: Sep 01, 2020 (05:30:52 UTC)

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(-)

Revision: 4b92804dc2fe6642a360
Author: Davis King
Date: Sep 01, 2020 (02:58:35 UTC)

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(-)

Revision: 0e721e5cae66214a57fa
Author: Davis King
Date: Aug 29, 2020 (05:09:54 UTC)

Fix bug in bounding box regression loss.

 dlib/dnn/loss.h | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

Revision: c9809e067f9e9c33aa1f
Author: Adrià Arrufat
Date: Aug 29, 2020 (21:04:24 UTC)

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(+)

Revision: b401185aa5a59bfff8eb
Author: Davis King
Date: Aug 23, 2020 (18:22:40 UTC)

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(-)

Revision: dd06c1169b55fc1f8994
Author: Adrià Arrufat
Date: Aug 24, 2020 (20:15:16 UTC)

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(+)

Revision: d7ca478b799762bae42c
Author: Juha Reunanen
Date: Aug 20, 2020 (17:43:14 UTC)

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(-)

Revision: bea99ceed046f5433c3f
Author: Davis King
Date: Aug 19, 2020 (15:48:14 UTC)

switch to a name less likely to conflict with third party code

 dlib/assert.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Revision: a9592b07fd0658491d1f
Author: Juha Reunanen
Date: Aug 20, 2020 (05:38:35 UTC)

Minor typo fixes (#2150)

 dlib/cuda/cudnn_dlibapi.cpp | 24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)

Revision: 2a870e329cbe6e551f6a
Author: samaldana
Date: Aug 20, 2020 (03:37:57 UTC)

Fix warning for zero variadic macro arguments. (#2151)

 dlib/assert.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Revision: f3b4fc548dd5fe83cdaf
Author: pfeatherstone
Date: Aug 18, 2020 (01:15:53 UTC)

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(-)

Revision: 59b44849bd00b00e37d9
Author: Davis King
Date: Aug 13, 2020 (03:47:59 UTC)

fix typo, doesn't really matter, but still

 dlib/global_optimization/find_max_global.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Revision: 02e70ce3cac806cfed20
Author: Davis King
Date: Aug 08, 2020 (11:30:37 UTC)

Record last changeset and set PATCH version to 99

 dlib/CMakeLists.txt | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Revision: 9117bd784328d9ac40ff
Author: Davis King
Date: Aug 08, 2020 (11:26:07 UTC)

Created release v19.21

 dlib/CMakeLists.txt | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)