{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "3725096f",
   "metadata": {},
   "source": [
    "# Compare Perceptual Similarity between two Videos with Feluda\n",
    "\n",
    "This notebook demonstrates how to use the `VideoHash` operator to generate\n",
    "perceptual hashes for two videos and compare their similarity. It processes\n",
    "two sample videos and shows their hash values for similarity analysis."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c32d1dd4",
   "metadata": {},
   "source": [
    "[![GitHub](https://img.shields.io/badge/GitHub-View%20Source-blue?logo=github)](https://github.com/tattle-made/feluda/blob/main/docs/examples/compare_video_hashes.ipynb) [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/tattle-made/feluda/blob/main/docs/examples/compare_video_hashes.ipynb)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b63d79e1",
   "metadata": {},
   "source": [
    "Install dependencies conditionally based on whether the notebook is running in Colab or locally."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "97d892e0",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running Notebook locally\n",
      "\u001b[2mUsing Python 3.10.12 environment at: /home/aatman/Aatman/Tattle/feluda/.venv\u001b[0m\n",
      "\u001b[2mAudited \u001b[1m6 packages\u001b[0m \u001b[2min 11ms\u001b[0m\u001b[0m\n",
      "CPU times: user 6.38 ms, sys: 4.13 ms, total: 10.5 ms\n",
      "Wall time: 138 ms\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "import sys\n",
    "\n",
    "IN_COLAB = \"google.colab\" in sys.modules\n",
    "print(\"Running Notebook in Google Colab\" if IN_COLAB else \"Running Notebook locally\")\n",
    "\n",
    "if IN_COLAB:\n",
    "    # Since Google Colab has preinstalled libraries like tensorflow and numba, we create a folder called feluda_custom_venv and isolate the environment there.\n",
    "    # This is done to avoid any conflicts with the preinstalled libraries.\n",
    "    %pip install uv\n",
    "    !mkdir -p /content/feluda_custom_venv\n",
    "    !uv pip install --target=/content/feluda_custom_venv --prerelease allow feluda \"feluda-video-hash-tmk\" > /dev/null 2>&1\n",
    "\n",
    "    sys.path.insert(0, \"/content/feluda_custom_venv\")\n",
    "else:\n",
    "    !uv pip install feluda \"feluda-video-hash-tmk\" > /dev/null 2>&1"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "35a99415-ca33-4395-8bf0-84b109bcd382",
   "metadata": {},
   "source": [
    "We'll use one operator for this example."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "2143fc54",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Downloading TMK binary to /var/folders/4p/bw6h5x8x1nb_17vsgfc12dz00000gn/T/tmk-hash-video\n"
     ]
    }
   ],
   "source": [
    "from feluda.factory import VideoFactory\n",
    "from feluda.operators import VideoHash\n",
    "\n",
    "hasher = VideoHash()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "85d10961-4a9f-4027-acfc-a280de6a0888",
   "metadata": {},
   "outputs": [],
   "source": [
    "VIDEO_URLS = [\n",
    "    \"https://github.com/tattle-made/feluda_datasets/blob/main/feluda-sample-media/en-speech.mp4\",\n",
    "    \"https://github.com/tattle-made/feluda_datasets/blob/main/feluda-sample-media/hi-speech.mp4\",\n",
    "]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5a92ecda-1b9b-40f2-b159-42e8b68520d9",
   "metadata": {},
   "source": [
    "In the below codeblock, we are computing the perceptual hash for two videos using the `feluda-video-hash-tmk` operator. The operator uses the compiled TMK Binary from Facebook Research."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ad07d1a7-5e74-4604-aab4-6f53bca59dcd",
   "metadata": {},
   "outputs": [],
   "source": [
    "hashes = []\n",
    "for i, video_url in enumerate(VIDEO_URLS, 1):\n",
    "    # Convert GitHub blob URL to CDN raw URL for direct download\n",
    "    raw_url = video_url.replace(\"/blob/\", \"/raw/\")\n",
    "\n",
    "    # Download video using VideoFactory\n",
    "    video_obj = VideoFactory.make_from_url(raw_url)\n",
    "    video_path = video_obj[\"path\"]\n",
    "\n",
    "    # Generate TMK hash for the video\n",
    "    # Returns Base64-encoded pure average feature vector\n",
    "    hash_value = hasher.run(video_path)\n",
    "    hashes.append(hash_value)\n",
    "\n",
    "    # Display hash information\n",
    "    print(f\"Video {i} URL: {video_url}\")\n",
    "    print(f\"TMK Hash: {hash_value[:50]}...\")  # Show first 50 characters\n",
    "    print(f\"Hash Length: {len(hash_value)} characters\")\n",
    "    print()\n",
    "\n",
    "# Compare the two hashes for similarity\n",
    "hash1, hash2 = hashes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e698bd28",
   "metadata": {},
   "outputs": [],
   "source": [
    "import base64\n",
    "\n",
    "# Decode Base64 hashes to compare raw feature vectors\n",
    "try:\n",
    "    decoded1 = base64.b64decode(hash1)\n",
    "    decoded2 = base64.b64decode(hash2)\n",
    "\n",
    "    # Calculate similarity metrics\n",
    "    hash_length = len(decoded1)\n",
    "    print(f\"Feature vector length: {hash_length} bytes\")\n",
    "\n",
    "    # Simple byte-by-byte comparison\n",
    "    identical_bytes = sum(1 for a, b in zip(decoded1, decoded2, strict=False) if a == b)\n",
    "    similarity_percentage = (identical_bytes / hash_length) * 100\n",
    "\n",
    "    print(f\"Identical bytes: {identical_bytes}/{hash_length}\")\n",
    "    print(f\"Similarity: {similarity_percentage:.2f}%\")\n",
    "\n",
    "    # Interpret similarity\n",
    "    if similarity_percentage > 80:\n",
    "        print(\"Result: HIGH SIMILARITY - Videos are likely very similar\")\n",
    "    elif similarity_percentage > 50:\n",
    "        print(\"Result: MODERATE SIMILARITY - Videos share some characteristics\")\n",
    "    else:\n",
    "        print(\"Result: LOW SIMILARITY - Videos are likely different\")\n",
    "\n",
    "except Exception as e:\n",
    "    print(f\"Error comparing hashes: {e}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "29a50a4a-a1f9-415f-9bfe-780c9427cab5",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Clean up resources when you're done\n",
    "\n",
    "hasher.cleanup()"
   ]
  }
 ],
 "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": "3.13.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}