add sheila

licenses/license-20190817-jarbas.txt

@@ -3,4 +3,5 @@ I grant anyone the right to use my work for any purpose, without any conditions,
 No warranties are given.
 
 Files covered by this update:
-marvin-en-0.3.0-20190817-jarbas.tar.gz
+marvin-en-0.3.0-20190817-jarbas.tar.gz
+sheila-en-0.3.0-20190817-jarbas.tar.gz

marvin/models/README.md

@@ -14,7 +14,6 @@ wake_words=1680 not_wake_words=102444 test_wake_words=420 test_not_wake_words=25
 
 {
     "tested_on": "test_set",
-    "name": "goog_marvin_e1000_s0.1_b100_mturk_synth_goog_pd",
     "tp": 362,
     "tn": 25592,
     "acc": 99.72,
@@ -24,7 +23,6 @@ wake_words=1680 not_wake_words=102444 test_wake_words=420 test_not_wake_words=25
 
 {
     "tested_on": "train_set",
-    "name": "goog_marvin_e1000_s0.2_b3000_mturk_synth_goog_pd",
     "tp": 1512,
     "tn": 102383,
     "acc": 99.84,

sheila/models/README.md

@@ -0,0 +1,33 @@
+# Sheila
+
+### sheila-en 20190817
+
+1000 epochs, sensitivity 0.1, batch_size 3000
+
+[Google Speech Commands Dataset ](https://ai.googleblog.com/2017/08/launching-speech-commands-dataset.html) "sheila" for ww samples
+
+[Public Domain Sounds Backup](http://pdsounds.tuxfamily.org/) + the other [Google Speech Commands Dataset ](https://ai.googleblog.com/2017/08/launching-speech-commands-dataset.html) + private dataset for nww.
+
+precise 0.3.0
+
+wake_words=1618 not_wake_words=102506 test_wake_words=404 test_not_wake_words=25635
+
+{
+    "tested_on": "test_set",
+    "tp": 356,
+    "tn": 25618,
+    "acc": 99.8,
+    "fn": 48,
+    "fp": 4
+}
+
+Tested on data used for training
+
+{
+    "tested_on": "train_set",
+    "tp": 1529,
+    "tn": 102444,
+    "acc": 99.91,
+    "fn": 89,
+    "fp": 1
+}