Somshubra Majumdar
b0d9591e39
Removed extra comment line
2017-03-03 19:05:45 -06:00
Somshubra Majumdar
6c656f12d9
Make SubPixelUpscaling work with any backend/dim ordering combination
2017-03-01 15:08:44 -06:00
Somshubra Majumdar
ee0e31a9b8
Fix typo
2017-03-01 14:08:01 -06:00
Somshubra Majumdar
55d565f327
Add documentation to SubPixelUpscaling layer
2017-03-01 14:04:01 -06:00
Somshubra Majumdar
96595f9b97
Merge branch 'master' of https://github.com/farizrahman4u/keras-contrib into densenet_fcn
2017-03-01 13:23:29 -06:00
Junwei Pan and GitHub
7e48f49dc3
Merge pull request #39 from tboquet/docfix
...
Doc and style fix
2017-03-01 08:31:29 -08:00
Junwei Pan and GitHub
90648786d0
Merge pull request #37 from farizrahman4u/cosineconv
...
Cosine Normalized Conv2D
2017-03-01 08:31:18 -08:00
Somshubra Majumdar
0bd720c6f3
Corrected a bug with naming in SubPixelUpsampling
2017-02-27 23:17:44 -06:00
Somshubra Majumdar
c40c680767
Correct name depth to space instead of depth to scale
2017-02-27 23:04:54 -06:00
Somshubra Majumdar
2e466d07d7
Revert pytest settings
2017-02-27 23:01:34 -06:00
Somshubra Majumdar
3634ca6bfc
Add assertions to prevent using wrong dim ordering/backends when using depth_to_space backend methods
2017-02-27 23:00:52 -06:00
Somshubra Majumdar
1986348465
Fix tests
2017-02-27 22:54:52 -06:00
Somshubra Majumdar
4aa331d548
Fix a few mistakes
2017-02-27 22:25:30 -06:00
Somshubra Majumdar
70507df3c6
Attempt running test again
2017-02-27 22:12:55 -06:00
Somshubra Majumdar
c86bb9364d
Added Subpixel convolution layer test
2017-02-27 21:13:58 -06:00
Michael Oliver
603e5188bb
pep8 fix
2017-02-27 15:06:47 -08:00
Michael Oliver
8db75c0619
move div to end for speed up
2017-02-27 14:19:52 -08:00
Somshubra Majumdar
fe020327c4
Correct the implementation to be more in line with the paper
2017-02-27 16:11:18 -06:00
tboquet
b86deb141f
* break line in several operations
2017-02-26 15:09:18 -05:00
tboquet
0eb0a0b08f
* doc and style fix
2017-02-26 14:39:31 -05:00
Michael Oliver
95ed5ad06d
clean up implementation
2017-02-22 21:14:37 -08:00
Michael Oliver
4d0368ad05
use consistent dim ordering
2017-02-22 20:41:20 -08:00
Michael Oliver
65db762520
make dim ordering match backend for test
2017-02-22 20:37:42 -08:00
Michael Oliver
03155630c9
remove filter reversal for tf test
2017-02-22 18:30:39 -08:00
Michael Oliver
d8ca85f445
fix tf dim ordering test
2017-02-22 18:23:39 -08:00
Michael Oliver
5b273df036
filter shape bug fix
2017-02-22 18:08:22 -08:00
Michael Oliver
36c0c9e600
PEP8 fixes
2017-02-22 16:47:38 -08:00
Michael Oliver
18600c2c67
fix pep8 test
2017-02-22 16:34:46 -08:00
Michael Oliver
7576c64985
fix pep8 test
2017-02-22 16:34:15 -08:00
Michael Oliver and GitHub
2b0092da2e
Update test_core.py
...
Make PEP8
2017-02-22 16:33:35 -08:00
Michael Oliver
de7b5eba5d
Merge branch 'master' into cosineconv
2017-02-22 16:31:07 -08:00
Michael Oliver
0e417813da
finish cosine conv2d
2017-02-22 16:30:17 -08:00
Michael Oliver and GitHub
8141f9ee22
Update to test for scale invariance
2017-02-22 15:04:54 -08:00
Michael Oliver and GitHub
4dc4821ff7
Cosine Normalized Dense Layer ( #36 )
...
* Add cosine normalized dense
* fixes and doc updates
* fix serialization
* fix >2D case
* change atol on allclose
2017-02-22 14:46:23 -08:00
Michael Oliver
af962bed2f
Merge branch 'cosine' into cosineconv
2017-02-22 13:07:12 -08:00
Michael Oliver
f996579d4c
initial commit
2017-02-22 13:06:52 -08:00
Michael Oliver
538b9cd45a
change atol on allclose
2017-02-22 11:59:38 -08:00
Michael Oliver
000927f7c9
fix >2D case
2017-02-22 11:45:42 -08:00
Michael Oliver
ac44e1a19c
fix serialization
2017-02-22 11:33:43 -08:00
Michael Oliver
82b92142cf
fixes and doc updates
2017-02-22 11:26:36 -08:00
Somshubra Majumdar
091a21dc12
Merge branch 'master' of https://github.com/farizrahman4u/keras-contrib into densenet_fcn
2017-02-22 13:20:43 -06:00
Michael Oliver
32536f79de
Merge branch 'master' into cosine
2017-02-22 11:07:35 -08:00
Michael Oliver
b10e9d795f
Add cosine normalized dense
2017-02-22 11:05:37 -08:00
Somshubra Majumdar and Michael Oliver
bdc64bf2f3
Added Batch Renormalization Layer ( #28 )
...
* Began work on batch renormalization
* Correct the implementation to use broadcasting properly, so it supports th dim ordering
* Correct the implementation to use broadcasting properly, so it supports th dim ordering
* Corrected normalizing term for d
* Moved location of updates
* Fixed the training problems and now auto adjusts the r_max and d_max values
* Removed unused code comments
* Finished implementation of batch renormalization
* Corrected the get_config method
* Added tests
* Added user changeable parameter t_delta to change the rate at which time steps increase
* Updated tests to state renorm everywhere
* Attempt a fix for mode=1 test fail (giving nans)
* Applied fix for nans to all calculations involving running mean and std
* Corrected time delta default value
* Reverts a mistake in pytest settings
* Corrections to calculations of r and d due to wrong variable naming (running_std perserves running_variance)
* Added t_delta to get_config
* Fix commit change mixup
* Corrected the check for uses_learning_flag
* Made a few corrections, added moments to backend, added test to backend
* Probable fix for test
* Corrected backend test
* Added support for mode 2
2017-02-22 09:07:13 -08:00
Somshubra Majumdar
968a0f4b84
Corrected changes as requested
2017-02-21 19:38:42 -06:00
Somshubra Majumdar
b84f748596
Added warning when using atrous conv upscaling and fall back to default
2017-02-21 12:30:01 -06:00
Somshubra Majumdar
bdd9ef593e
Added proper support for changing parameter nb_layers_per_block in original DenseNet
2017-02-21 12:15:25 -06:00
Somshubra Majumdar
84511789e5
Fix PEP8
2017-02-21 09:51:18 -06:00
Somshubra Majumdar
3b2b5eca77
Corrected some comments
2017-02-21 09:50:21 -06:00
Somshubra Majumdar
8f9d387265
Managed to reduce the code duplication by shifting the builder to densenet.py
2017-02-21 01:25:43 -06:00
Somshubra Majumdar