test/test_clap_score.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import sys\n",
    "import shutil\n",
    "import numpy as np\n",
    "import soundfile as sf\n",
    "\n",
    "module_path = os.path.abspath(os.path.join('..'))\n",
    "if module_path not in sys.path:\n",
    "    sys.path.append(module_path)\n",
    "\n",
    "from frechet_audio_distance import CLAPScore\n",
    "from utils import gen_sine_wave"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 630k-audioset (for general audio less than 10-sec)\n",
    "SAMPLE_RATE = 48000\n",
    "LENGTH_IN_SECONDS = 2\n",
    "\n",
    "# init\n",
    "clap = CLAPScore(\n",
    "    ckpt_dir=\"../checkpoints/clap\",\n",
    "    submodel_name=\"630k-audioset\",\n",
    "    verbose=True,\n",
    "    audio_load_worker=8,\n",
    "    enable_fusion=False,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# TEST USING SINE WAVES\n",
    "\n",
    "# generate sine waves and text\n",
    "text_audio1 = []\n",
    "text_audio2 = []\n",
    "\n",
    "# text and audio 1 (sine waves between 100-1000 Hz)\n",
    "text_folder1, audio_folder1, count, noise_param = \"text1\", \"audio1\", 10, None\n",
    "os.makedirs(text_folder1, exist_ok=True)\n",
    "os.makedirs(audio_folder1, exist_ok=True)\n",
    "frequencies = np.linspace(100, 1000, count).tolist()\n",
    "for freq in frequencies:\n",
    "    # text\n",
    "    text_audio1.append(f\"sin_{freq:.0f}.wav,{freq:.0f} Hz sine wave\")\n",
    "    # audio\n",
    "    samples = gen_sine_wave(freq, LENGTH_IN_SECONDS, SAMPLE_RATE, param=noise_param)\n",
    "    filename = os.path.join(audio_folder1, \"sin_%.0f.wav\" % freq)\n",
    "    print(\"Creating: %s with %i samples.\" % (filename, samples.shape[0]))\n",
    "    sf.write(filename, samples, SAMPLE_RATE, \"PCM_24\")\n",
    "# save text to csv\n",
    "with open(os.path.join(text_folder1, \"text.csv\"), \"w\") as f:\n",
    "    f.write(\"filename,caption\\n\")\n",
    "    for line in text_audio1:\n",
    "        f.write(line + \"\\n\")\n",
    "\n",
    "# text and audio 2 (noisy sine waves between 1000-10000 Hz)\n",
    "text_folder2, audio_folder2, count, noise_param = \"text2\", \"audio2\", 10, 0.5\n",
    "os.makedirs(text_folder2, exist_ok=True)\n",
    "os.makedirs(audio_folder2, exist_ok=True)\n",
    "frequencies = np.linspace(1000, 10000, count).tolist()\n",
    "for freq in frequencies:\n",
    "    # text\n",
    "    text_audio2.append(f\"sin_{freq:.0f}.wav,{freq:.0f} Hz sine wave\")\n",
    "    # audio\n",
    "    samples = gen_sine_wave(freq, LENGTH_IN_SECONDS, SAMPLE_RATE, param=noise_param)\n",
    "    filename = os.path.join(audio_folder2, \"sin_%.0f.wav\" % freq)\n",
    "    print(\"Creating: %s with %i samples.\" % (filename, samples.shape[0]))\n",
    "    sf.write(filename, samples, SAMPLE_RATE, \"PCM_24\")\n",
    "# save text to csv\n",
    "with open(os.path.join(text_folder2, \"text.csv\"), \"w\") as f:\n",
    "    f.write(\"filename,caption\\n\")\n",
    "    for line in text_audio2:\n",
    "        f.write(line + \"\\n\")\n",
    "\n",
    "# text 1 vs. audio 1\n",
    "clap_score = clap.score(\n",
    "    text_path=\"./text1/text.csv\",\n",
    "    audio_dir=\"./audio1\",\n",
    "    text_column=\"caption\",\n",
    "    text_embds_path=None,\n",
    "    audio_embds_path=None,\n",
    ")\n",
    "print(f\"CLAP score text 1 vs audio 1 [mu. std]: {clap_score}\")\n",
    "\n",
    "# text 2 vs. audio 2\n",
    "clap_score = clap.score(\n",
    "    text_path=\"./text2/text.csv\",\n",
    "    audio_dir=\"./audio2\",\n",
    "    text_column=\"caption\",\n",
    "    text_embds_path=None,\n",
    "    audio_embds_path=None,\n",
    ")\n",
    "print(f\"CLAP score text 2 vs audio 2 [mu. std]: {clap_score}\")\n",
    "\n",
    "# text 1 vs. audio 2\n",
    "clap_score = clap.score(\n",
    "    text_path=\"./text1/text.csv\",\n",
    "    audio_dir=\"./audio2\",\n",
    "    text_column=\"caption\",\n",
    "    text_embds_path=None,\n",
    "    audio_embds_path=None,\n",
    ")\n",
    "print(f\"CLAP score text 1 vs audio 2 [mu. std]: {clap_score}\")\n",
    "\n",
    "# text 2 vs. audio 1\n",
    "clap_score = clap.score(\n",
    "    text_path=\"./text2/text.csv\",\n",
    "    audio_dir=\"./audio1\",\n",
    "    text_column=\"caption\",\n",
    "    text_embds_path=None,\n",
    "    audio_embds_path=None,\n",
    ")\n",
    "print(f\"CLAP score text 2 vs audio 1 [mu. std]: {clap_score}\")\n",
    "\n",
    "shutil.rmtree(\"text1\")\n",
    "shutil.rmtree(\"audio1\")\n",
    "shutil.rmtree(\"text2\")\n",
    "shutil.rmtree(\"audio2\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# text and audio matching\n",
    "clap_score = clap.score(\n",
    "    text_path=\"./clap_score_text/text_matching.csv\",\n",
    "    audio_dir=\"./clap_score_audio\",\n",
    "    text_column=\"caption\",\n",
    "    text_embds_path=None,\n",
    "    audio_embds_path=None,\n",
    ")\n",
    "print(f\"CLAP score text and audio matching [mu. std]: {clap_score}\")\n",
    "\n",
    "# text and audio not matching\n",
    "clap_score = clap.score(\n",
    "    text_path=\"./clap_score_text/text_not_matching.csv\",\n",
    "    audio_dir=\"./clap_score_audio\",\n",
    "    text_column=\"caption\",\n",
    "    text_embds_path=None,\n",
    "    audio_embds_path=None,\n",
    ")\n",
    "print(f\"CLAP score text and audio not matching [mu. std]: {clap_score}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": ".venv",
   "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": "3.9.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}