Commit Graph

7842 Commits

Author SHA1 Message Date
Davis King cc0cbc0248 Fixed typo in travis build script. 2019-03-17 21:33:28 -04:00
Davis King 3401b96936 It's actually fine to call rowm() or colm() with an empty index set. 2019-03-17 21:30:49 -04:00
Davis King ec63baa7a2 Have travis build some tests with DLIB_ASSERT enabled to make sure they work. 2019-03-17 21:15:08 -04:00
Davis King f8827f6bb3 Make test deterministic 2019-03-17 09:06:09 -04:00
Davis King 89fb5c2e31 cleanup 2019-03-15 08:46:03 -04:00
Davis King 4e4acf794f Make CMake build a test program to see if libjpeg is really available and not broken. 2019-03-15 08:44:49 -04:00
Davis King 7611b34768 Set png include paths for test. 2019-03-15 08:28:12 -04:00
Davis King 0cf422d42a Have CMake build a little test program to see if the system copy of libpng really works. 2019-03-15 08:23:36 -04:00
Davis King 0268e27481 updated docs 2019-03-14 07:59:22 -04:00
Davis King e72d5c71f1 Exlude unneeded files from python source distributions. 2019-03-10 11:46:44 -04:00
Davis King 1682c83e85 Removed old mercurial ignore file. 2019-03-10 11:36:54 -04:00
Davis King 579a41d523 Record last changeset and set PATCH version to 99 2019-03-10 11:09:53 -04:00
Davis King 810ef7937d Created release v19.17 2019-03-10 11:04:56 -04:00
Davis King 590a9a11b1 updated docs 2019-03-10 11:01:50 -04:00
Davis King 55d5750286 updated docs 2019-03-10 10:48:53 -04:00
Davis King 533e7a3740 updated release notes 2019-03-10 10:23:50 -04:00
Davis King ffce4af6e2 Make cusolver finding work for really old versions of cmake with CUDA 10.1 2019-03-10 10:02:29 -04:00
Davis King 5faf8ccce3 Automatically test for NEON instructions when building python extensions and enable them if available.
Forgot about --yes USE_NEON_INSTRUCTIONS when removing --yes a moment
ago.  It's on by default now so now --yes really is unneeded.  The dlib
policy on this kind of thing going forward is to always have a cmake
script that tests if a feature like this is available and to enable it
automatically when building python extensions.
2019-03-08 08:05:11 -05:00
Davis King b892df8232 Removed --yes option from setup.py since it has long been a noop and its presence just confuses users. 2019-03-08 07:51:33 -05:00
Davis King 83385e2356 Added more dianostic messages from cmake. 2019-03-08 07:35:45 -05:00
Davis King 663a1b7127 Fixed more gcc 8 warnings. 2019-03-03 21:35:15 -05:00
Davis King f9b7ad9ae6 Fixed compiler warnings in gcc 8 2019-03-03 21:17:20 -05:00
Davis King 5d03b99a08 Changed to avoid compiler warning. 2019-03-03 20:12:43 -05:00
Davis King 02ed083c4c Work around bug in nvcc in cuda 10.1 2019-03-03 17:01:48 -05:00
Davis King c3ee638b3f update ignore list 2019-03-03 17:01:38 -05:00
Davis King ba206da012 Switched doc making scripts to pull info from git instead of hg. 2019-03-03 11:22:25 -05:00
Davis King 70f9a9f11f Made cusolver finding work in newer versions of cmake and cuda 2019-03-02 09:45:08 -05:00
Kapil Sachdeva f7f6f67618 Add python api that generates desciptor(s) from the aligned image(s) (#1667)
* Add python api that generates desciptor(s) from the aligned image(s)

* Remove asserts from face_recognition.py example/tutorial

* In batch_compute_face_descriptors_from_aligned_images, use for-in loop to simplify the code

Improvde the document on binding methods and the error message if the aligned image is not of size 150x150
2019-02-25 20:10:04 -05:00
Juha Reunanen 04a2387cfc Make sure CUDA_VERSION is defined (#1663) 2019-02-20 08:33:01 -05:00
Davis King 23f874d6e3 Fixed pytest not working with python 2.7. So turning travis tests back on for that version of python. 2019-02-20 08:30:38 -05:00
Davis King 23c140d138 Removing the python 2.7 test on travis because there is suddenly some kind of
incompatibility between python 2.7 and pytest, but only on travis.
2019-02-20 08:02:44 -05:00
Davis King 16a1056916 Removing the install_requires cmake from setup.py because the pip cmake package is hopelessly busted at the moment. 2019-02-20 07:54:42 -05:00
Davis King 12ca8ad67f merged 2019-02-20 07:48:50 -05:00
Davis King 9e18e26205 merged 2019-02-20 07:48:13 -05:00
Davis King 12646178ee updated docs 2019-02-18 07:59:14 -05:00
Branko Kokanovic 41a87e5926 Exposing chinese_whispers directly, closes #1642 (#1644)
* Exposing chinese_whispers directly

* Addressing comments

* Changed description
* Added support for distance weights
* Added tests

* Improving test to check returned results
2019-02-07 06:59:04 -08:00
davemers0160 84b72278b5 fixed check for excessive detections in loss_mmod_ (#1625)
fixed check for excessive detections in loss_mmod_

Ran into the problem where dets.size() was equal to max_num_initial_dets which then throws a subscript out of range error when accesing: dets[max_num_initial_dets].detection_confidence.  This fixes that issue.
2019-01-19 20:45:46 -05:00
Davis King ea45199572 Remove Python 3.5 tests on AppVeyor because something is wrong with python 3.5 on appveyor and the other tests are sufficient for python coverage. 2019-01-19 20:40:21 -05:00
Davis King ab9ce99846 merged 2019-01-09 09:08:24 -05:00
Davis King 027e7faf4e Add cmake as a pip install dependency. 2019-01-09 09:07:52 -05:00
Juha Reunanen 5c623e3c93 Apply the #1514 fix even on non-Windows platforms, and change all remaining cudaStreamSynchronize calls (#1596)
* Apply the #1514 fix even on non-Windows platforms

* Try to fix #1513 even more by circumventing the remaining cudaStreamSynchronize calls

* Make the fix apply only if CUDA_VERSION == 10000

* Make the fix apply on CUDA 9.2 also

* CHECK_CUDA(cudaStreamSynchronize(stream));
2019-01-07 22:43:34 -05:00
Chris Mayo 9ca72b3e1c Reliably detect AVX2 at runtime on Linux (#1614)
ECX must be set to 0 to read this flag, the easiest way to achieve this
is to use __cpuid_count with 0 as the second argument.
2019-01-07 22:41:26 -05:00
Davis King 6b5b461dd0 updated docs 2019-01-06 15:38:11 -05:00
Davis King be8dc9269b merged 2019-01-06 15:34:04 -05:00
Davis King c763fafde9 Fixed broken memcpy overload. 2019-01-06 15:32:20 -05:00
Juha Reunanen f685cb4249 Add U-net style skip connections to the semantic-segmentation example (#1600)
* Add concat_prev layer, and U-net example for semantic segmentation

* Allow to supply mini-batch size as command-line parameter

* Decrease default mini-batch size from 30 to 24

* Resize t1, if needed

* Use DenseNet-style blocks instead of residual learning

* Increase default mini-batch size to 50

* Increase default mini-batch size from 50 to 60

* Resize even during the backward step, if needed

* Use resize_bilinear_gradient for the backward step

* Fix function call ambiguity problem

* Clear destination before adding gradient

* Works OK-ish

* Add more U-tags

* Tweak default mini-batch size

* Define a simpler network when using Microsoft Visual C++ compiler; clean up the DenseNet stuff (leaving it for a later PR)

* Decrease default mini-batch size from 24 to 23

* Define separate dnn filename for MSVC++ and not

* Add documentation for the resize_to_prev layer; move the implementation so that it comes after mult_prev

* Fix previous typo

* Minor formatting changes

* Reverse the ordering of levels

* Increase the learning-rate stopping criterion back to 1e-4 (was 1e-8)

* Use more U-tags even on Windows

* Minor formatting

* Latest MSVC 2017 builds fast, so there's no need to limit the depth any longer

* Tweak default mini-batch size again

* Even though latest MSVC can now build the extra layers, it does not mean we should add them!

* Fix naming
2019-01-06 09:11:39 -05:00
Thomas Peters fb4c62cc67 remove register keyword from jdhuff for C++17 (#1612)
the register storage specifier was deprecated in C++11 and removed in C++17
2019-01-04 22:38:55 -05:00
Facundo Galán 61a021c932 Add pointwise_divide function (#1586)
* Add pointwise_divide operator

* Add tests for pointwise_divide function

* Replace in affine layer pointwise_multiply for division by its equivalent pointwise_divide.
2018-12-11 07:22:14 -05:00
Davis King 99af7b9d54 Fixed code not compiling. 2018-12-05 21:41:34 -05:00
RedfalconAI 4b2ef5b8f9 Update CmakeLists: add png for arm (#1580) 2018-12-05 21:36:50 -05:00