YAMNet variable sr (#8174)

* Modify yamnet_visualization to work for any (?) sampling rate in input wav file.
tensorflow · Feb 22, 2020 · 7b1553f · 7b1553f
1 parent b60dc23
commit 7b1553f
Showing 1 changed file with 28 additions and 15 deletions.
diff --git a/research/audioset/yamnet/yamnet_visualization.ipynb b/research/audioset/yamnet/yamnet_visualization.ipynb
@@ -53,6 +53,34 @@
    "cell_type": "code",
    "execution_count": 2,
    "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Sample rate = 16000\n"
+     ]
+    }
+   ],
+   "source": [
+    "# Read in the audio.\n",
+    "# You can get this example waveform via:\n",
+    "# curl -O https://storage.googleapis.com/audioset/speech_whistling2.wav\n",
+    "\n",
+    "wav_file_name = 'speech_whistling2.wav'\n",
+    "\n",
+    "wav_data, sr = sf.read(wav_file_name, dtype=np.int16)\n",
+    "waveform = wav_data / 32768.0\n",
+    "# The graph is designed for a sampling rate of 16 kHz, but higher rates \n",
+    "# should work too.\n",
+    "params.SAMPLE_RATE = sr\n",
+    "print(\"Sample rate =\", params.SAMPLE_RATE)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
    "outputs": [
     {
      "name": "stdout",
@@ -74,21 +102,6 @@
     "    yamnet.load_weights('yamnet.h5')"
    ]
   },
-  {
-   "cell_type": "code",
-   "execution_count": 3,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "# Read in the audio.\n",
-    "# You can get this example waveform via:\n",
-    "# curl -O https://storage.googleapis.com/audioset/speech_whistling2.wav\n",
-    "wav_data, sr = sf.read('speech_whistling2.wav', dtype=np.int16)\n",
-    "waveform = wav_data / 32768.0\n",
-    "# Sampling rate should be 16000 Hz.\n",
-    "assert sr == 16000"
-   ]
-  },
   {
    "cell_type": "code",
    "execution_count": 4,