AudioClassification.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "!pip install tensorflow  tensorflow-io matplotlib"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import os\n",
    "from matplotlib import pyplot as plt\n",
    "import tensorflow as tf \n",
    "import tensorflow_io as tfio"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "CAPUCHIN_FILE = os.path.join('data', 'Parsed_Capuchinbird_Clips', 'XC3776-3.wav')\n",
    "NOT_CAPUCHIN_FILE = os.path.join('data', 'Parsed_Not_Capuchinbird_Clips', 'afternoon-birds-song-in-forest-0.wav')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "def load_wav_16k_mono(filename):\n",
    "    # Load encoded wav file\n",
    "    file_contents = tf.io.read_file(filename)\n",
    "    # Decode wav (tensors by channels) \n",
    "    wav, sample_rate = tf.audio.decode_wav(file_contents, desired_channels=1)\n",
    "    # Removes trailing axis\n",
    "    wav = tf.squeeze(wav, axis=-1)\n",
    "    sample_rate = tf.cast(sample_rate, dtype=tf.int64)\n",
    "    # Goes from 44100Hz to 16000hz - amplitude of the audio signal\n",
    "    wav = tfio.audio.resample(wav, rate_in=sample_rate, rate_out=16000)\n",
    "    return wav"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-01-29 18:30:39.864912: W tensorflow/core/framework/op_kernel.cc:1839] OP_REQUIRES failed at whole_file_read_ops.cc:116 : NOT_FOUND: data/Parsed_Capuchinbird_Clips/XC3776-3.wav; No such file or directory\n"
     ]
    },
    {
     "ename": "NotFoundError",
     "evalue": "{{function_node __wrapped__ReadFile_device_/job:localhost/replica:0/task:0/device:CPU:0}} data/Parsed_Capuchinbird_Clips/XC3776-3.wav; No such file or directory [Op:ReadFile]",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNotFoundError\u001b[0m                             Traceback (most recent call last)",
      "Cell \u001b[0;32mIn[7], line 1\u001b[0m\n\u001b[0;32m----> 1\u001b[0m wave \u001b[38;5;241m=\u001b[39m load_wav_16k_mono(CAPUCHIN_FILE)\n\u001b[1;32m      2\u001b[0m nwave \u001b[38;5;241m=\u001b[39m load_wav_16k_mono(NOT_CAPUCHIN_FILE)\n",
      "Cell \u001b[0;32mIn[6], line 3\u001b[0m, in \u001b[0;36mload_wav_16k_mono\u001b[0;34m(filename)\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21mload_wav_16k_mono\u001b[39m(filename):\n\u001b[1;32m      2\u001b[0m     \u001b[38;5;66;03m# Load encoded wav file\u001b[39;00m\n\u001b[0;32m----> 3\u001b[0m     file_contents \u001b[38;5;241m=\u001b[39m tf\u001b[38;5;241m.\u001b[39mio\u001b[38;5;241m.\u001b[39mread_file(filename)\n\u001b[1;32m      4\u001b[0m     \u001b[38;5;66;03m# Decode wav (tensors by channels) \u001b[39;00m\n\u001b[1;32m      5\u001b[0m     wav, sample_rate \u001b[38;5;241m=\u001b[39m tf\u001b[38;5;241m.\u001b[39maudio\u001b[38;5;241m.\u001b[39mdecode_wav(file_contents, desired_channels\u001b[38;5;241m=\u001b[39m\u001b[38;5;241m1\u001b[39m)\n",
      "File \u001b[0;32m~/anaconda3/lib/python3.11/site-packages/tensorflow/python/ops/io_ops.py:134\u001b[0m, in \u001b[0;36mread_file\u001b[0;34m(filename, name)\u001b[0m\n\u001b[1;32m     97\u001b[0m \u001b[38;5;129m@tf_export\u001b[39m(\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mio.read_file\u001b[39m\u001b[38;5;124m\"\u001b[39m, v1\u001b[38;5;241m=\u001b[39m[\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mio.read_file\u001b[39m\u001b[38;5;124m\"\u001b[39m, \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mread_file\u001b[39m\u001b[38;5;124m\"\u001b[39m])\n\u001b[1;32m     98\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21mread_file\u001b[39m(filename, name\u001b[38;5;241m=\u001b[39m\u001b[38;5;28;01mNone\u001b[39;00m):\n\u001b[1;32m     99\u001b[0m \u001b[38;5;250m  \u001b[39m\u001b[38;5;124;03m\"\"\"Reads the contents of file.\u001b[39;00m\n\u001b[1;32m    100\u001b[0m \n\u001b[1;32m    101\u001b[0m \u001b[38;5;124;03m  This operation returns a tensor with the entire contents of the input\u001b[39;00m\n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m    132\u001b[0m \u001b[38;5;124;03m    A tensor of dtype \"string\", with the file contents.\u001b[39;00m\n\u001b[1;32m    133\u001b[0m \u001b[38;5;124;03m  \"\"\"\u001b[39;00m\n\u001b[0;32m--> 134\u001b[0m   \u001b[38;5;28;01mreturn\u001b[39;00m gen_io_ops\u001b[38;5;241m.\u001b[39mread_file(filename, name)\n",
      "File \u001b[0;32m~/anaconda3/lib/python3.11/site-packages/tensorflow/python/ops/gen_io_ops.py:583\u001b[0m, in \u001b[0;36mread_file\u001b[0;34m(filename, name)\u001b[0m\n\u001b[1;32m    581\u001b[0m   \u001b[38;5;28;01mpass\u001b[39;00m\n\u001b[1;32m    582\u001b[0m \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[0;32m--> 583\u001b[0m   \u001b[38;5;28;01mreturn\u001b[39;00m read_file_eager_fallback(\n\u001b[1;32m    584\u001b[0m       filename, name\u001b[38;5;241m=\u001b[39mname, ctx\u001b[38;5;241m=\u001b[39m_ctx)\n\u001b[1;32m    585\u001b[0m \u001b[38;5;28;01mexcept\u001b[39;00m _core\u001b[38;5;241m.\u001b[39m_SymbolicException:\n\u001b[1;32m    586\u001b[0m   \u001b[38;5;28;01mpass\u001b[39;00m  \u001b[38;5;66;03m# Add nodes to the TensorFlow graph.\u001b[39;00m\n",
      "File \u001b[0;32m~/anaconda3/lib/python3.11/site-packages/tensorflow/python/ops/gen_io_ops.py:606\u001b[0m, in \u001b[0;36mread_file_eager_fallback\u001b[0;34m(filename, name, ctx)\u001b[0m\n\u001b[1;32m    604\u001b[0m _inputs_flat \u001b[38;5;241m=\u001b[39m [filename]\n\u001b[1;32m    605\u001b[0m _attrs \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;01mNone\u001b[39;00m\n\u001b[0;32m--> 606\u001b[0m _result \u001b[38;5;241m=\u001b[39m _execute\u001b[38;5;241m.\u001b[39mexecute(\u001b[38;5;124mb\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mReadFile\u001b[39m\u001b[38;5;124m\"\u001b[39m, \u001b[38;5;241m1\u001b[39m, inputs\u001b[38;5;241m=\u001b[39m_inputs_flat,\n\u001b[1;32m    607\u001b[0m                            attrs\u001b[38;5;241m=\u001b[39m_attrs, ctx\u001b[38;5;241m=\u001b[39mctx, name\u001b[38;5;241m=\u001b[39mname)\n\u001b[1;32m    608\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m _execute\u001b[38;5;241m.\u001b[39mmust_record_gradient():\n\u001b[1;32m    609\u001b[0m   _execute\u001b[38;5;241m.\u001b[39mrecord_gradient(\n\u001b[1;32m    610\u001b[0m       \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mReadFile\u001b[39m\u001b[38;5;124m\"\u001b[39m, _inputs_flat, _attrs, _result)\n",
      "File \u001b[0;32m~/anaconda3/lib/python3.11/site-packages/tensorflow/python/eager/execute.py:53\u001b[0m, in \u001b[0;36mquick_execute\u001b[0;34m(op_name, num_outputs, inputs, attrs, ctx, name)\u001b[0m\n\u001b[1;32m     51\u001b[0m \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[1;32m     52\u001b[0m   ctx\u001b[38;5;241m.\u001b[39mensure_initialized()\n\u001b[0;32m---> 53\u001b[0m   tensors \u001b[38;5;241m=\u001b[39m pywrap_tfe\u001b[38;5;241m.\u001b[39mTFE_Py_Execute(ctx\u001b[38;5;241m.\u001b[39m_handle, device_name, op_name,\n\u001b[1;32m     54\u001b[0m                                       inputs, attrs, num_outputs)\n\u001b[1;32m     55\u001b[0m \u001b[38;5;28;01mexcept\u001b[39;00m core\u001b[38;5;241m.\u001b[39m_NotOkStatusException \u001b[38;5;28;01mas\u001b[39;00m e:\n\u001b[1;32m     56\u001b[0m   \u001b[38;5;28;01mif\u001b[39;00m name \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m:\n",
      "\u001b[0;31mNotFoundError\u001b[0m: {{function_node __wrapped__ReadFile_device_/job:localhost/replica:0/task:0/device:CPU:0}} data/Parsed_Capuchinbird_Clips/XC3776-3.wav; No such file or directory [Op:ReadFile]"
     ]
    }
   ],
   "source": [
    "wave = load_wav_16k_mono(CAPUCHIN_FILE)\n",
    "nwave = load_wav_16k_mono(NOT_CAPUCHIN_FILE)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "plt.plot(wave)\n",
    "plt.plot(nwave)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "POS = os.path.join('data', 'Parsed_Capuchinbird_Clips')\n",
    "NEG = os.path.join('data', 'Parsed_Not_Capuchinbird_Clips')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "pos = tf.data.Dataset.list_files(POS+'\\*.wav')\n",
    "neg = tf.data.Dataset.list_files(NEG+'\\*.wav')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "positives = tf.data.Dataset.zip((pos, tf.data.Dataset.from_tensor_slices(tf.ones(len(pos)))))\n",
    "negatives = tf.data.Dataset.zip((neg, tf.data.Dataset.from_tensor_slices(tf.zeros(len(neg)))))\n",
    "data = positives.concatenate(negatives)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "lengths = []\n",
    "for file in os.listdir(os.path.join('data', 'Parsed_Capuchinbird_Clips')):\n",
    "    tensor_wave = load_wav_16k_mono(os.path.join('data', 'Parsed_Capuchinbird_Clips', file))\n",
    "    lengths.append(len(tensor_wave))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "tf.math.reduce_mean(lengths)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "tf.math.reduce_min(lengths)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "tf.math.reduce_max(lengths)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "def preprocess(file_path, label): \n",
    "    wav = load_wav_16k_mono(file_path)\n",
    "    wav = wav[:48000]\n",
    "    zero_padding = tf.zeros([48000] - tf.shape(wav), dtype=tf.float32)\n",
    "    wav = tf.concat([zero_padding, wav],0)\n",
    "    spectrogram = tf.signal.stft(wav, frame_length=320, frame_step=32)\n",
    "    spectrogram = tf.abs(spectrogram)\n",
    "    spectrogram = tf.expand_dims(spectrogram, axis=2)\n",
    "    return spectrogram, label"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "filepath, label = positives.shuffle(buffer_size=10000).as_numpy_iterator().next()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "spectrogram, label = preprocess(filepath, label)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "plt.figure(figsize=(30,20))\n",
    "plt.imshow(tf.transpose(spectrogram)[0])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "data = data.map(preprocess)\n",
    "data = data.cache()\n",
    "data = data.shuffle(buffer_size=1000)\n",
    "data = data.batch(16)\n",
    "data = data.prefetch(8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "train = data.take(36)\n",
    "test = data.skip(36).take(15)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "samples, labels = train.as_numpy_iterator().next()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "samples.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Conv2D, Dense, Flatten"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "model = Sequential()\n",
    "model.add(Conv2D(16, (3,3), activation='relu', input_shape=(1491, 257,1)))\n",
    "model.add(Conv2D(16, (3,3), activation='relu'))\n",
    "model.add(Flatten())\n",
    "model.add(Dense(128, activation='relu'))\n",
    "model.add(Dense(1, activation='sigmoid'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "model.compile('Adam', loss='BinaryCrossentropy', metrics=[tf.keras.metrics.Recall(),tf.keras.metrics.Precision()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "hist = model.fit(train, epochs=4, validation_data=test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "plt.title('Loss')\n",
    "plt.plot(hist.history['loss'], 'r')\n",
    "plt.plot(hist.history['val_loss'], 'b')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "plt.title('Precision')\n",
    "plt.plot(hist.history['precision'], 'r')\n",
    "plt.plot(hist.history['val_precision'], 'b')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "plt.title('Recall')\n",
    "plt.plot(hist.history['recall'], 'r')\n",
    "plt.plot(hist.history['val_recall'], 'b')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "X_test, y_test = test.as_numpy_iterator().next()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "yhat = model.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "yhat = [1 if prediction > 0.5 else 0 for prediction in yhat]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    }
   },
   "outputs": [],
   "source": [
    "def load_mp3_16k_mono(filename):\n",
    "    \"\"\" Load a WAV file, convert it to a float tensor, resample to 16 kHz single-channel audio. \"\"\"\n",
    "    res = tfio.audio.AudioIOTensor(filename)\n",
    "    # Convert to tensor and combine channels \n",
    "    tensor = res.to_tensor()\n",
    "    tensor = tf.math.reduce_sum(tensor, axis=1) / 2 \n",
    "    # Extract sample rate and cast\n",
    "    sample_rate = res.rate\n",
    "    sample_rate = tf.cast(sample_rate, dtype=tf.int64)\n",
    "    # Resample to 16 kHz\n",
    "    wav = tfio.audio.resample(tensor, rate_in=sample_rate, rate_out=16000)\n",
    "    return wav"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    }
   },
   "outputs": [],
   "source": [
    "mp3 = os.path.join('data', 'Forest Recordings', 'recording_00.mp3')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "wav = load_mp3_16k_mono(mp3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    }
   },
   "outputs": [],
   "source": [
    "audio_slices = tf.keras.utils.timeseries_dataset_from_array(wav, wav, sequence_length=48000, sequence_stride=48000, batch_size=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    }
   },
   "outputs": [],
   "source": [
    "samples, index = audio_slices.as_numpy_iterator().next()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    }
   },
   "outputs": [],
   "source": [
    "def preprocess_mp3(sample, index):\n",
    "    sample = sample[0]\n",
    "    zero_padding = tf.zeros([48000] - tf.shape(sample), dtype=tf.float32)\n",
    "    wav = tf.concat([zero_padding, sample],0)\n",
    "    spectrogram = tf.signal.stft(wav, frame_length=320, frame_step=32)\n",
    "    spectrogram = tf.abs(spectrogram)\n",
    "    spectrogram = tf.expand_dims(spectrogram, axis=2)\n",
    "    return spectrogram"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    }
   },
   "outputs": [],
   "source": [
    "audio_slices = tf.keras.utils.timeseries_dataset_from_array(wav, wav, sequence_length=16000, sequence_stride=16000, batch_size=1)\n",
    "audio_slices = audio_slices.map(preprocess_mp3)\n",
    "audio_slices = audio_slices.batch(64)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    }
   },
   "outputs": [],
   "source": [
    "yhat = model.predict(audio_slices)\n",
    "yhat = [1 if prediction > 0.5 else 0 for prediction in yhat]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    }
   },
   "outputs": [],
   "source": [
    "from itertools import groupby"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    }
   },
   "outputs": [],
   "source": [
    "yhat = [key for key, group in groupby(yhat)]\n",
    "calls = tf.math.reduce_sum(yhat).numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    }
   },
   "outputs": [],
   "source": [
    "calls"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    }
   },
   "outputs": [],
   "source": [
    "results = {}\n",
    "for file in os.listdir(os.path.join('data', 'Forest Recordings')):\n",
    "    FILEPATH = os.path.join('data','Forest Recordings', file)\n",
    "    \n",
    "    wav = load_mp3_16k_mono(FILEPATH)\n",
    "    audio_slices = tf.keras.utils.timeseries_dataset_from_array(wav, wav, sequence_length=48000, sequence_stride=48000, batch_size=1)\n",
    "    audio_slices = audio_slices.map(preprocess_mp3)\n",
    "    audio_slices = audio_slices.batch(64)\n",
    "    \n",
    "    yhat = model.predict(audio_slices)\n",
    "    \n",
    "    results[file] = yhat"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "scrolled": true,
    "tags": []
   },
   "outputs": [],
   "source": [
    "results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "scrolled": true,
    "tags": []
   },
   "outputs": [],
   "source": [
    "class_preds = {}\n",
    "for file, logits in results.items():\n",
    "    class_preds[file] = [1 if prediction > 0.99 else 0 for prediction in logits]\n",
    "class_preds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "scrolled": true,
    "tags": []
   },
   "outputs": [],
   "source": [
    "postprocessed = {}\n",
    "for file, scores in class_preds.items():\n",
    "    postprocessed[file] = tf.math.reduce_sum([key for key, group in groupby(scores)]).numpy()\n",
    "postprocessed"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    }
   },
   "outputs": [],
   "source": [
    "import csv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "with open('results.csv', 'w', newline='') as f:\n",
    "    writer = csv.writer(f, delimiter=',')\n",
    "    writer.writerow(['recording', 'capuchin_calls'])\n",
    "    for key, value in postprocessed.items():\n",
    "        writer.writerow([key, value])"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}