7d49bef758
Fix pep8 issues.
2015-12-21 19:09:54 -05:00
eecb0041e1
Improvements to create-train-val-split.
...
+ Use python2 for consistency with our other Python code.
+ The previous version left empty subdirectories.
This modification deletes them.
2015-12-21 18:08:39 -05:00
e90694fcba
nn4: Add LRN and remove 5x5 convolutions for #55 .
...
Reduces the number of parameters from 7472144 to 6959088.
2015-12-20 22:29:55 -05:00
d69cd17d13
Docs: Remove typo.
2015-12-20 17:36:32 -05:00
027f9dcf8e
Real-Time Web Demo: Simplify server initialization.
2015-12-19 18:17:27 -05:00
7d73946836
Web demo: server.py -> websocket-server.py
2015-12-19 17:46:57 -05:00
aec7de78dc
Explictly set txaio.use_twisted.
...
https://groups.google.com/forum/#!topic/cmu-openface/zbWEuYCf7_w
2015-12-18 10:52:36 -05:00
918d3af141
Add port option to server.
2015-12-18 10:52:36 -05:00
eea992b458
Mention Python dependencies for training.
2015-12-17 11:43:00 -05:00
33c2b687ed
Link to cunn.
2015-12-17 09:09:41 -05:00
92672ec406
Merge.
2015-12-17 09:06:46 -05:00
bf38abe8f9
Remove unused cudnn dependency.
2015-12-17 09:04:46 -05:00
dff471987e
Link to NVIDIA cuDNN.
2015-12-17 08:53:00 -05:00
4a0449e44c
Correct cudnn package name.
2015-12-17 08:46:17 -05:00
fe56897ddc
Fix pep8 warning.
2015-12-16 13:14:39 -05:00
163f06d32b
Add hack for CentOS support for #66 .
2015-12-16 13:12:15 -05:00
1b9ad376ff
Add timing information to classifier.py
...
Thanks @lucafeudi
2015-12-14 11:53:15 -05:00
139cfb7756
tsne: Print where file is saved to.
2015-12-12 16:30:18 -05:00
446436ced1
align-dlib: Check if image was correctly loaded.
2015-12-12 15:27:54 -05:00
244ac78e90
Formatting.
2015-12-12 13:45:54 -05:00
af575bdb72
Update docs with new align-dlib interface.
2015-12-12 13:43:30 -05:00
b9c848c8a2
Fix pep8 warning.
2015-12-11 15:59:28 -05:00
c8a4f34095
Untested improvements to create-unknown-vectors.
2015-12-11 15:56:00 -05:00
97cb800a40
Fix pep8 errors.
2015-12-11 10:22:18 -05:00
ec7650cf73
Update server for the slightly modified alignment API.
2015-12-11 10:21:01 -05:00
0e99d0c499
Merge branch 'experimental/align'
2015-12-10 21:09:01 -05:00
12a8648959
Update demos with results from the new models.
2015-12-10 21:05:55 -05:00
4a94b51314
Add annotate-image util.
2015-12-10 21:05:55 -05:00
d81f76e48a
Update LFW results with changes from #50 .
2015-12-10 21:04:00 -05:00
05a1b87c6f
Mention fblualib dependency in docs.
2015-12-08 18:31:11 -05:00
08986266ea
Correct names for #50 .
2015-12-08 17:42:27 -05:00
d6809b119e
Update align-dlib with new alignment code for #50 .
2015-12-08 17:04:57 -05:00
e5fedb125c
Update align-dlib with new alignment code for #50 .
2015-12-08 17:04:21 -05:00
a1a89ea65f
Remove skimage dependency for #50 .
2015-12-08 16:41:40 -05:00
a78043923b
Alignment improvements for #50 .
...
Thanks @hbredin for the great feedback!
1. Make it clear dlib expects RGB images.
2. Keep the mean landmarks in the alignment source code
rather than loading them from a txt file.
3. Normalize the mean landmarks between 0 and 1 for
the transformation.
2015-12-08 16:29:09 -05:00
7fd4480161
Docs: faq: Formatting.
2015-12-06 17:23:44 -05:00
b5d77344ca
Add improved alignment code for #50 . Accuracy not yet tested.
2015-12-06 16:46:25 -05:00
4962a4876d
Docs: Initial commit of FAQ.
2015-12-06 16:20:25 -05:00
0a24569354
Move repo contents from docs to README.
2015-12-05 09:09:02 -05:00
0701e69db8
Docs: Link to accuracy page.
2015-12-04 03:14:37 -05:00
5e6977c8cd
Classifier demo: Check the mode before checking classifierModel.
...
For #65 .
2015-12-03 13:47:37 -05:00
0efe4fe391
Docs: Setup: Add note about Docker memory.
2015-12-03 13:40:35 -05:00
6742ef326f
Docs: Update links to other projects in 'accuracy'
2015-12-02 03:28:53 -05:00
cf49f815c6
docs: Remove semicolon from for loop for bash
2015-12-01 16:29:52 -05:00
44e5a6401d
Improve web demo docs.
2015-11-30 17:33:27 -05:00
98002816f8
Merge pull request #61 from ZhihengCV/master
...
solve the memory leak problem of web-demo
2015-11-28 00:21:38 -05:00
334461d34f
solve the memory leak problem of web-demo
2015-11-28 11:13:40 +08:00
09577c5d06
Automatically switch video feeds.
2015-11-26 15:21:18 -05:00
f1f4f63598
Web Demo: Show fiducial points and use tabs instead of scolling.
2015-11-26 13:31:27 -05:00
c48255a236
Add better error handling when the Torch model is passed as the classification model.
2015-11-24 09:57:33 -05:00
Brandon Amos