{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Predict\n",
    "Similar to training, prediction can be done via three interfaces:\n",
    "- via python, `das.predict.predict`\n",
    "- via the command line, `das predict`, with audio data from a wav file.\n",
    "- the GUI - see the [GUI tutorial](/tutorials_gui/predict)\n",
    "\n",
    "Prediction will:\n",
    "\n",
    "- load the audio data and the network\n",
    "- run inference to produce confidence scores (`class_probabilties`)\n",
    "- post-process the confidence score to extract the times of events and label segments.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Prediction using python"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from pprint import pprint\n",
    "import scipy.io.wavfile\n",
    "import das.predict\n",
    "help(das.predict.predict)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "%%time\n",
    "samplerate, x = scipy.io.wavfile.read('dat/dmel_song_rt.wav')\n",
    "print(f\"DAS requires [T, channels], but single-channel wave files are loaded with shape [T,] (data shape is {x.shape}).\")\n",
    "x = np.atleast_2d(x).T\n",
    "events, segments, class_probabilities, class_names = das.predict.predict(x, \n",
    "                                                           model_save_name='models/dmel_single_rt/20200430_201821',\n",
    "                                                           verbose=2,\n",
    "                                                           segment_minlen=0.02,\n",
    "                                                           segment_fillgap=0.02)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Outputs of `predict`\n",
    "- `class_probabilties`: `[T, nb_classes]` including noise.\n",
    "- `segments`: Labelled segments\n",
    "    - `samplerate_Hz`: \n",
    "    - `names`: names of all segment types\n",
    "    - `index`: indices of all segments types into class_probabiltiies\n",
    "    - `probabilities = class_probabilites[:, index]`\n",
    "    - `sequence`: sequence of segment names (one entry per detected segment). Excludes noise\n",
    "    - `samples`: labelled sample trace (label of the sequence occupying each sample)\n",
    "    - `onsets_seconds`, `offsets_seconds`, `durations_seconds`: Onsets, offsets, and duration of individual segmeents\n",
    "- `events`: Detected events\n",
    "    - `samplerate_Hz`: \n",
    "    - `index`: indices of all events types into class_probabiltiies\n",
    "    - `names`: names of all event types\n",
    "    - `probabilities`: probabilities (confidence scores) for detected events. Value of `class_probabilities` for the detected event index at each event time.\n",
    "    - `seconds`: times (seconds) of detected events\n",
    "    - `sequence`: sequence of event names (one per detected event)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "plt.style.use('ncb.mplstyle')\n",
    "\n",
    "t0 = 0\n",
    "t1 = 30_000 \n",
    "fs =segments['samplerate_Hz']\n",
    "time = np.arange(t0, t1) / fs\n",
    "nb_classes = class_probabilities.shape[1]\n",
    "\n",
    "plt.figure(figsize=(30, 10))\n",
    "plt.subplot(411)\n",
    "plt.plot(time, x[t0:t1], 'k', linewidth=0.5)\n",
    "plt.title('Song')\n",
    "plt.xticks([])\n",
    "plt.ylim(-0.25, 0.25)\n",
    "\n",
    "plt.subplot(412)\n",
    "plt.imshow(class_probabilities[t0:t1].T, cmap='Greys')\n",
    "plt.yticks(np.arange(nb_classes), labels=class_names)\n",
    "plt.title('Raw confidence scores')\n",
    "plt.xticks([])\n",
    "\n",
    "ax = plt.subplot(413)\n",
    "plt.plot(time, x[t0:t1],'k', linewidth=0.5)\n",
    "plt.ylim(-0.25, 0.25)\n",
    "plt.title('Annotations')\n",
    "plt.xlabel('Time [seconds]')\n",
    "for onset, offset, segment_name in zip(segments['onsets_seconds'], segments['offsets_seconds'], segments['sequence']):\n",
    "    if onset >= t0 /fs and offset <= t1 / fs:\n",
    "        plt.plot([onset, offset], [0.1, 0.1], c='b')\n",
    "        ax.annotate(segment_name, xy=(onset, 0.11), c='b')\n",
    "\n",
    "for pulse_time, pulse_name in zip(events['seconds'], events['sequence']):\n",
    "    if pulse_time >= t0 /fs and pulse_time <= t1 / fs:\n",
    "        plt.axvline(pulse_time, c='r')\n",
    "        ax.annotate(pulse_name, xy=(pulse_time, 0.1), c='r', rotation=-90)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Prediction using command-line scripts\n",
    "Will save the output of `das.predict.predict` to a h5 file ending in `_das.h5` or specified via the `--save-filename` argument.\n",
    "\n",
    "See [cli](/technical/cli) for a full list of arguments."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "!das predict dat/dmel_song_rt.wav models/dmel_single_rt/20200430_201821"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import h5py\n",
    "with h5py.File('dat/dmel_song_rt_das.h5', mode='r') as f:\n",
    "    print(list(f.keys()))"
   ]
  }
 ],
 "metadata": {
  "file_extension": ".py",
  "kernelspec": {
   "display_name": "Python 3.9.9 ('base')",
   "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": "3.9.9"
  },
  "mimetype": "text/x-python",
  "name": "python",
  "npconvert_exporter": "python",
  "pygments_lexer": "ipython3",
  "version": 3,
  "vscode": {
   "interpreter": {
    "hash": "7ea0ec616133ead53c1908c8f6539f5c0cb9b2f78368e2bb6ab3f847e89ca400"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}