Commit Graph

6809 Commits

Author SHA1 Message Date
Davis King faf8676a49 Cleanup 2017-10-27 05:58:34 -04:00
Davis King e8d74b3015 Removed old cruft not needed anymore since we are doing this kind
of thing with cmake targets already in set_compiler_specific_options.cmake.
2017-10-27 05:50:15 -04:00
Sean Warren 16ad749c45 Win lapack (#913)
* Fall back on find_package for blas, lapack on Windows

* Remove debugging message
2017-10-27 05:45:53 -04:00
Davis King aa93c8f861 Updated python code to use the new dlib::jitter_image() instead of hacking it
out of the random_cropper.
2017-10-25 05:42:31 -04:00
Davis King e338bf02e0 Changed the random_cropper's set_min_object_size() routine to take min box
dimensions in the same format as the mmod_options object (i.e. two lengths
measured in pixels).  This should make defining random_cropping strategies that
are consistent with MMOD settings much more straightforward since you can just
take the mmod_options settings and give them to the random_cropper and it will
do the right thing.
2017-10-24 22:10:02 -04:00
Davis King 1c664eeac5 Made the metric learning example do image jittering. 2017-10-24 21:13:02 -04:00
Davis King 369f2b32e8 Cleaned up jitter_image() code and moved it into dlib proper. 2017-10-24 08:02:44 -04:00
Davis King 782f4f4825 merged 2017-10-23 21:02:57 -04:00
Sean Warren 45ce9f7b6d Use banded Cholesky factorization if possible (#857)
* Use banded Cholesky factorization if possible
Computation cost from n.n.n -> n.n.b where b is band size

* Tidy up whitespace

* Remove typo

*  Escape from banded matrix detection correctly

* Use LAPACK banded Cholesky factorisation where possible

* Add banded chol tests

* Add test for banded chol in column major layout

* Use row major layout for banded chol - more efficient as we will pass to LAPACK
2017-10-23 21:00:49 -04:00
Davis King 04a8f0093d Make sure the test loss the trainer logs to the console never gets suck at
infinity.
2017-10-22 16:06:11 -04:00
Gilles Rochefort 9bc7070a77 Add some operator() to cv_image for compatibility with mmod loss. (#900)
* Add some operator() to cv_image for compatibility with mmod.

* Update documentation
2017-10-21 10:16:36 -04:00
Davis King 261f12d4ea Updated spec 2017-10-20 21:45:14 -04:00
Gilles Rochefort 540f47409e Missing interfaces in add_prev for compatibility with mmod loss. (#901) 2017-10-20 21:44:00 -04:00
Davis King 6d343f93da Sometimes the loss_mmod_ layer could experience excessively long runtime during
early iterations since the model might produce a huge number of false alarms
while the detector is still bad.  Processing all these detections can cause it
to run slowly until the model is good enough to avoid really excessive amounts
of false alarms.  This change puts more of a limit on the number of false
alarms processed during those early iterations and avoids the slowdown.
2017-10-20 21:37:54 -04:00
Davis King 3e48a36ede The loss returned by compute_loss_value_and_gradient() wasn't quite right. It
doesn't effect normal use, but it's still wrong and this change fixes it.
2017-10-20 21:34:34 -04:00
Davis King 3da61b2581 Added install_target_output_folder option to matlab cmake scripts. 2017-10-20 16:54:49 -04:00
Davis King 1d8445cbfd Fixed the terrible breaking change to the tensor object I made earlier today :( 2017-10-19 20:57:24 -04:00
Davis King 86465e307a Improved cmake error messages about CUDA support. 2017-10-19 12:09:29 -04:00
Davis King e9837f7035 Changed tensor so that, when reallocating memory, it frees any existing memory
*before* allocating new memory.  It used to be the other way around which
caused momentary spikes of increased memory usage.  This could put you over the
total memory available in some cases which is obviously less than ideal
behavior.
2017-10-19 10:50:40 -04:00
Davis King 7a317f5456 Fixed bug 2017-10-18 10:17:46 -04:00
Davis King e35b2d8fc8 Added python binary classifier example 2017-10-18 07:36:31 -04:00
Davis King 525bf0e15b Fixed weird wording. 2017-10-17 19:20:29 -04:00
Davis King 3aa9991892 Made comment more clear. 2017-10-17 19:18:57 -04:00
Juha Reunanen 9d59214062 Add get_net parameter that allows to call the function without forcing flush to disk (#875)
* Add get_net parameter allowing to call the function without forced flush to disk (see the discussion in #869)

* A blindfolded attempt to fix compile error on the CI server
2017-10-17 19:12:38 -04:00
jpblackburn f467236cdd Allow cross_validate_trainer_threaded to use non-double data (#883) 2017-10-17 19:11:16 -04:00
Davis King 2b9b3fef6a merged 2017-10-17 19:02:58 -04:00
Davis King aa1a3b8611 Made cmake print a very explicit error message if you try to use a version of gcc that is too old. 2017-10-17 19:02:31 -04:00
Davis King 88d0bdec86 Fixed path 2017-10-17 10:31:27 -04:00
Davis King 50c1099449 Yet more cmake cleanup 2017-10-17 10:24:23 -04:00
Davis King d832a5760a More cmake script cleanup 2017-10-17 10:02:30 -04:00
Davis King e17b082540 Fix deprecation warning from new cmake 2017-10-17 09:45:18 -04:00
Davis King 38a2846dda More cmake cleanup 2017-10-16 21:36:18 -04:00
Davis King 6ffdf802ee suppress compiler warnings 2017-10-16 21:34:17 -04:00
Davis King 5b4206f97c Suppress compiler warnings 2017-10-16 21:34:00 -04:00
Davis King f18f753c04 CMake tweaks to avoid errors in visual studio 2017-10-16 21:11:16 -04:00
Davis King 3b2d73db68 Some cmake script cleanup and refactoring. 2017-10-16 20:55:28 -04:00
Davis King c0b60faaf0 Removed cout statement. 2017-10-16 20:55:07 -04:00
Davis King 0282b9428c Fixed compiler error due to #define nonsense in visual studio. 2017-10-16 09:22:01 -04:00
Davis King 1dc5e00299 Prevent include path clashes from system installs of dlib as well as suppress
warning messages from boost and python system header files.
2017-10-16 06:49:03 -04:00
Davis King 363c1ad963 merged 2017-10-15 15:54:19 -04:00
Davis King 08a7483724 I accidentally slightly changed the cropping behavior of
get_face_chip_details() when used with the 68 point landmark model about a
month ago.  This change reverts it back to the previous behavior.  The change
was very minor, so it shouldn't matter either way.  But being consistent is
important and I'm changing it back.
2017-10-15 15:53:33 -04:00
Davis King a21093d392 Added warning messages when users use the old non-target based cmake variables. 2017-10-14 10:29:29 -04:00
Davis King 221ad16d3c merged 2017-10-10 09:59:59 -04:00
Davis King 79a620b18e Added assert to improve error messages. 2017-10-10 09:53:22 -04:00
Davis King 8365bf6878 Added get_integer() and get_integer_in_range() to dlib::rand. 2017-10-10 09:52:43 -04:00
Davis King 38dd831d9f Made dlib::array able to push_back() from rvalues. 2017-10-08 07:49:50 -04:00
Davis King 17a6e682e5 Fixed missing inline 2017-10-08 07:45:50 -04:00
Davis King 40448e64bd Made add_image_left_right_flips() and add_image_rotations() work with
mmod_rects in addition to rectangles and full_object_detections.
2017-10-07 14:09:09 -04:00
Davis King 4470280946 Fixed load_image_dataset()'s skip_empty_images() option. It wasn't skipping
images that only have ignore boxes when you load into mmod_rects like it should
have been.
2017-10-07 14:05:58 -04:00
Davis King c2a9dee846 merged 2017-10-04 16:22:19 -04:00