06430cd24d
Add --freeze-till training parameter for transfer learning
...
For example, --freeze-till -2 will freeze layers[0:-2]
2019-04-09 02:09:06 -05:00
f870827116
Cast to np.array in Stats
...
This fixes an error with precise-graph
2019-04-09 02:07:42 -05:00
d3920f36dd
Add threshold parameters and precise-calc-threshold script
...
This is used to produce a more linear output
2019-04-08 15:08:16 -05:00
18ab3d1e1d
Refine graph visualization on data file
2019-04-03 05:48:18 -05:00
b3f2ff8a2d
Add option to print stats to file
2019-04-03 05:14:57 -05:00
b4c28e1771
Drastically improve precise-graph speed
...
It now uses numpy arrays instead of Python loops
2019-04-03 04:42:42 -05:00
d609d783ad
Fix data loading on empty directory
2019-03-29 05:23:49 -07:00
baf08cd7ee
Add pyache data loading
...
This enables loading the cache as a block without risking filling up the harddrive
2019-03-29 06:40:14 -05:00
66eed19660
Move model creation before data loading
...
This allows it to error if the model isn't correct before loading data which is less annoying
2019-03-29 06:00:55 -05:00
e1e68ac7ab
Fix divide by 0 in Stats
2019-03-29 05:58:32 -05:00
608c9b43f7
Add threshold to precise-train-incremental
2019-03-29 02:58:00 -05:00
8b9d003b5b
Update requirements.txt
2019-03-29 01:50:11 -05:00
1d5e002e80
Upgrade tensorflow to 1.13.x
2019-03-29 01:48:43 -05:00
61d52929e5
Make precise-eval use new stats object
...
Fixes import error
2019-03-21 17:45:09 -05:00
00ca1fc683
Add threshold command line option to scripts
...
Also makes the term threshold consistent in code
2019-03-20 23:36:08 -05:00
9453f8a2d1
Add more options to graph script
2019-03-20 23:12:10 -05:00
63e5553129
Use new Stats class in precise-test
2019-03-20 22:33:59 -05:00
3373dc81bb
Add curl to the dependencies
...
It is needed to download pip and is not installed by default in docker
2019-03-20 14:42:17 -05:00
4ee464bdb5
Fix stats bug
2019-03-20 11:38:16 -05:00
4cd1927efe
Update copyright header to 2019
2019-03-20 10:53:24 -05:00
a8a7cea1c6
Add missing method docstring
2019-03-20 10:52:25 -05:00
e89677a84f
Add precise-train-generated script
2019-03-20 10:45:39 -05:00
2d52d533a1
Add update_vectors method to Listener
...
This can be used to use the vectors from the listener without a network runner
2019-03-20 00:06:10 -05:00
02da99e21d
Add precise-graph script to show ROC curves
2019-03-20 00:05:59 -05:00
c1b6677f48
Add pytest cache to .gitignore
2019-03-18 13:54:26 -05:00
4cb695da4d
Upgrade requirements.txt
2019-01-27 07:33:25 +00:00
530f76bffb
Upgrade tensorflow to 1.11.0
2019-01-26 23:23:04 +00:00
904bd7168e
Add exception_on_overflow=False to pyaudio.Stream.read
...
This resolves some issues with reading streams on certain platforms (#59 )
This also removes the read_divisor in favor of a wrapper for pyaudio which slightly simplifies the code
2019-01-26 16:54:20 -06:00
4284fc83ae
Update README
...
- Add runner.start() to sample scripts
- Specify only Linux Desktops are supported in the Binary install
2019-01-26 14:09:51 -06:00
2b13175b9f
Add link to pretrained models in README
2019-01-21 14:31:56 -06:00
a1ad138bf0
Re-add speechpy to requirements ( Fixes #45 )
...
This prevents older models from erroring
2018-11-11 21:23:01 -06:00
7d539f34da
Make sourcing the venv clearer
2018-11-07 16:27:49 -06:00
3827d08a48
Add extra binary install readme info ( Fixes #40 )
2018-11-05 20:56:19 -06:00
e3c74dd44d
Improve export script
2018-10-19 12:55:31 -05:00
692df01e01
Make shuffling configurable ( Fixes #28 )
...
This fixes filenames in precise-test being incorrect
2018-10-15 14:12:40 -05:00
8b6e05ebc0
Fix exception on Ubuntu
2018-10-12 18:50:09 -05:00
6b34a58339
Update export script to auto package tar and calculate md5
2018-10-12 11:04:08 -05:00
da286a2dd7
Make noise ratio configurable via command line
2018-10-09 19:07:48 -05:00
78603bd3c2
Add precise-add-noise script
...
This is used to generate noise into a dataset
2018-09-05 05:02:52 -05:00
33f0e48167
Prevent corrupt group json from stopping training
...
It can be regenerated so it's not a big deal if it gets corrupt
2018-09-05 05:01:21 -05:00
90d8671cf7
Fix delay samples in train-incremental
...
Also fixes bug with model parameters
2018-09-05 01:53:07 -05:00
b3e6efaf53
Fix failing to load params
...
Fixes #26
2018-08-11 10:12:34 -05:00
91184444cb
Remove broken tensorflow download code
...
For the Raspberry Pi, it is now officially supported. It also would incorrectly run on desktop python 3.4 installations
2018-08-10 01:02:33 -05:00
80b13db2e7
Improve formatting line break
2018-08-10 00:03:16 -05:00
02fa053653
Fix loss bias in trainer.py
2018-08-10 00:02:39 -05:00
63b4592675
Fix activations_per_day in precise-simulate
2018-08-10 00:01:16 -05:00
e43498d2ea
Explicitly delete large vectors
...
This reduces potential memory overlap
2018-08-10 00:00:30 -05:00
b1d7a7ff5d
Show percentage when calculating vectors
2018-08-09 23:59:26 -05:00
7bded13de4
Change vectorizer to Sonopy
...
Also adds vectorizer parameter to choose between mels and mfccs
2018-08-09 23:58:21 -05:00
754d332b10
Reorganize and improve setup.py
...
Added keywords and license keys
2018-08-09 23:10:59 -05:00
Matthew D. Scholefield