Davis King
f5574434db
Upgraded loss_mmod_ to support objects of varying aspect ratio. This changes
...
the API for the mmod_options struct slightly.
2017-07-01 14:46:27 -04:00
Davis King
dcfff1c463
Added process() and process_batch() to add_loss_layer. These routines let you
...
easily pass arguments to any optional parameters of a loss layer's to_tensor()
routine. For instance, it makes it more convenient to set loss_mmod_'s
adjust_threshold parameter.
2017-07-01 12:12:30 -04:00
Davis King
78103a582c
updated spec about con_
2017-07-01 11:54:20 -04:00
Davis King
8b805c99d9
updated docs
2017-07-01 11:53:53 -04:00
Davis King
b377f752d9
Made it so you can set the number of output filters for con_ layers at runtime.
2017-07-01 11:53:38 -04:00
Davis King
8eb9e295ee
Made dnn_trainer sync its state to two separate sync files that it alternates
...
between. This should make syncing more robust to sudden hardware failure that
happens right when saving to disk.
2017-07-01 11:41:26 -04:00
Davis King
a06b5292bd
Added select_oldest_file() and select_newest_file()
2017-07-01 11:37:16 -04:00
Davis King
0796ce79b4
updated docs
2017-07-01 11:34:35 -04:00
Davis King
2f4e373b21
Added file::last_modified() for windows version of code.
2017-07-01 11:20:45 -04:00
Davis King
e00352aa09
merged
2017-07-01 10:39:05 -04:00
Davis King
56428eb9c0
Added last_modified() method to dlib::file.
2017-07-01 10:38:15 -04:00
Davis King
023e13982d
merged
2017-06-28 08:55:47 -04:00
Davis King
e44da57705
Minor change to avoid errors from some versions of cmake.
2017-06-28 08:53:43 -04:00
Davis King
c68af9dcbf
Merge branch 'OranjeeGeneral-master'
2017-06-26 21:08:55 -04:00
Davis King
31bcddd5e4
Cleaned up documentation for conv_. Also removed unnecessary tensor
...
reallocation and copying inside conv_'s backward pass. Doing this
required adding an add_to_output boolean option to the methods of
tensor_conv.
2017-06-26 21:06:59 -04:00
Davis King
b3d5dbd3b3
Fixed typo in comment.
2017-06-26 21:01:47 -04:00
Davis King
f9bb4f472b
Merge branch 'master' of git://github.com/OranjeeGeneral/dlib into OranjeeGeneral-master
2017-06-26 14:59:03 -04:00
OranjeeGeneral
ecb7095e4a
refactored interface to reduce complexity so conv and convt layers forward passes have to call setup explicit now and there is only one ()-operator
2017-06-22 17:55:20 +01:00
Davis King
48c68f218e
Fixed mex class code printing
2017-06-21 17:06:48 -04:00
Davis King
c5847374f4
Added operator= for simd8f so assignment from float compiles.
2017-06-21 09:10:34 -04:00
Davis King
cbd187fb61
merged
2017-06-19 20:55:51 -04:00
Davis King
39be45ada2
Made is so pressing e in imglab toggles between views of the image where the
...
histogram is equalized or unmodified. This way, if you are looking at
particularly dark or badly contrasted images you can toggle this mode and maybe
get a better view of what you are labeling.
2017-06-19 20:54:45 -04:00
Davis King
ba72c2f95c
Updated code to work with new random_cropper interface.
2017-06-18 08:11:54 -04:00
Davis King
17b48b97bb
Changed the random_cropper's interface so that instead of talking in terms of
...
min and max object height, it's now min and max object size. This way, if you
have objects that are short and wide (i.e. objects where the relevant dimension
is width rather than height) you will get sensible behavior out of the random
cropper.
2017-06-17 12:34:26 -04:00
Joachim
3a91295ea7
Merge branch 'master' of https://github.com/davisking/dlib
2017-06-13 13:33:42 +01:00
Grigoris
75f6658223
Minor typo corrections in shape predictor trainer. ( #633 )
2017-06-07 06:51:19 -04:00
Plumtus
362bec1099
Reinitialize averagers when saved sync file was reloaded. ( #629 )
2017-06-06 14:19:23 -04:00
Davis King
d2b80bfe6f
Switched order of things in if statement so cmake hopefully won't give weird errors.
2017-06-06 05:47:00 -04:00
Davis King
74fbca45ca
Changed the converter so that, rather than producing one python file with
...
everything in it, it now makes a python file as before but an additional binary
file with all the weights in it. This way, if you are working with a network
with a very large number of weights you won't end up with a crazy large python
script.
2017-06-04 10:06:44 -04:00
Davis King
f9eab48813
updated docs
2017-06-02 22:51:59 -04:00
Davis King
3ee460da93
Added set_rect_area()
2017-06-02 22:50:01 -04:00
Davis King
5bc1792d4f
Added a .fill() member to curand_generator that can create random 32bit
...
integers.
2017-05-31 16:08:00 -04:00
Davis King
2ee1036299
merged
2017-05-30 17:18:49 -04:00
Davis King
e8e064e534
Added --compiler-flags to setup.py so you can pass options directly to gcc.
2017-05-30 17:18:04 -04:00
Evgeniy Fominov
9ed2ba9e5a
Possible CLang fix for Neon-based SIMD4i ( #612 )
2017-05-30 05:40:11 -04:00
Davis King
0ef3b736fd
Relaxed the default non-max suppression parameters used by the mmod_options
...
object so that users of the deep learning MMOD tool don't get spurious errors
about impossibly labeled objects during training.
2017-05-29 20:06:37 -04:00
Davis King
88383a848b
Made the converter handle caffe's odd pooling layer output size calculations.
2017-05-28 11:07:02 -04:00
Davis King
724cb5006f
Work around a bug in visual studio 2015.
2017-05-27 12:56:24 -04:00
Davis King
df19361c8f
Made calling clean() on network objects also call clean on any layer detail
...
objects that also provide a .clean() method.
2017-05-27 11:59:32 -04:00
Davis King
115e8b6dfa
updated docs
2017-05-26 17:24:44 -04:00
Davis King
e1f2bb3859
Added visit_layers_until_tag()
2017-05-26 17:23:58 -04:00
Davis King
44387e396d
merged
2017-05-24 07:25:51 -04:00
Davis King
a88f1bd8f2
Made the converter add zero padding layers when needed by Eltwise to replicate
...
the behavior of dlib's add_prev layers.
2017-05-24 07:24:12 -04:00
Davis King
984b694962
Made error message slightly better.
2017-05-22 19:11:00 -04:00
Davis King
cbda2b9e33
Changed caffe converter to require the user to specify the input tensor size
...
when the converter is run.
2017-05-22 19:06:55 -04:00
Juha Reunanen
97ff8cb2a9
Noticed compiler warning C4800: 'double': forcing value to bool 'true' or 'false' (performance warning) ( #538 )
...
- suggesting to change the test to what perhaps was the original intention
2017-05-22 08:58:27 -04:00
Davis King
523e020e18
updated docs
2017-05-20 09:40:48 -04:00
emptyVoid
465e95424d
Move shape_predictor_trainer to a separate header ( #599 )
...
Moved shape_predictor_trainer to its own header in order to support the
use of shape_predictor with DLIB_ISO_CPP_ONLY defined (i.e. without
threading API wrappers).
2017-05-20 09:37:40 -04:00
Davis King
10d3f93333
Dlib and caffe actually do use max pooling layers with padding in the same way.
...
So I just removed the error check that was preventing the conversion from
proceeding in that case. I also added more useful output messages about
setting input tensor dimensions.
2017-05-19 20:16:40 -04:00
Davis King
e28768eafa
Fixed uninitialized variable in test that caused unreliable test results.
2017-05-17 16:43:32 -04:00
