Segmentation_Depth_MLP.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "gpuType": "T4",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/1kaiser/Media-Segment-Depth-MLP/blob/main/Segmentation_Depth_MLP.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## ***RUNNING MEDIAPIPE HOLISTIC***"
      ],
      "metadata": {
        "id": "-juv_T5TX5Kp"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "!wget -nc -P /content/ --output-document=test.mp4 https://github.com/1kaiser/Media-Segment-Depth-MLP/releases/download/v0.2/video.mp4"
      ],
      "metadata": {
        "id": "K-EszVm8pR_h",
        "outputId": "65dfc334-70b8-4028-9bf5-01396b4d8a77",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "--2023-12-07 00:32:54--  https://github.com/1kaiser/Media-Segment-Depth-MLP/releases/download/v0.2/video.mp4\n",
            "Resolving github.com (github.com)... 140.82.114.4\n",
            "Connecting to github.com (github.com)|140.82.114.4|:443... connected.\n",
            "HTTP request sent, awaiting response... 302 Found\n",
            "Location: https://objects.githubusercontent.com/github-production-release-asset-2e65be/578313074/82c445be-7ee0-45ad-89d7-c22b3840f23e?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=AKIAIWNJYAX4CSVEH53A%2F20231207%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20231207T003254Z&X-Amz-Expires=300&X-Amz-Signature=756848cbb8d25c88a63c30a4759aee8ddb1a611114754eb9b8a1c9b285cbeece&X-Amz-SignedHeaders=host&actor_id=0&key_id=0&repo_id=578313074&response-content-disposition=attachment%3B%20filename%3Dvideo.mp4&response-content-type=application%2Foctet-stream [following]\n",
            "--2023-12-07 00:32:55--  https://objects.githubusercontent.com/github-production-release-asset-2e65be/578313074/82c445be-7ee0-45ad-89d7-c22b3840f23e?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=AKIAIWNJYAX4CSVEH53A%2F20231207%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20231207T003254Z&X-Amz-Expires=300&X-Amz-Signature=756848cbb8d25c88a63c30a4759aee8ddb1a611114754eb9b8a1c9b285cbeece&X-Amz-SignedHeaders=host&actor_id=0&key_id=0&repo_id=578313074&response-content-disposition=attachment%3B%20filename%3Dvideo.mp4&response-content-type=application%2Foctet-stream\n",
            "Resolving objects.githubusercontent.com (objects.githubusercontent.com)... 185.199.108.133, 185.199.109.133, 185.199.110.133, ...\n",
            "Connecting to objects.githubusercontent.com (objects.githubusercontent.com)|185.199.108.133|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 3415073 (3.3M) [application/octet-stream]\n",
            "Saving to: ‘test.mp4’\n",
            "\n",
            "                    100%[===================>]   3.26M  --.-KB/s    in 0.07s   \n",
            "\n",
            "2023-12-07 00:32:55 (49.9 MB/s) - ‘test.mp4’ saved [3415073/3415073]\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "!python -m pip install mediapipe"
      ],
      "metadata": {
        "id": "BUW-yGjEYGey",
        "outputId": "c9b50fcb-2925-4188-de75-f21aaf8b4720",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Collecting mediapipe\n",
            "  Downloading mediapipe-0.10.8-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (34.5 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m34.5/34.5 MB\u001b[0m \u001b[31m37.9 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hRequirement already satisfied: absl-py in /usr/local/lib/python3.10/dist-packages (from mediapipe) (1.4.0)\n",
            "Requirement already satisfied: attrs>=19.1.0 in /usr/local/lib/python3.10/dist-packages (from mediapipe) (23.1.0)\n",
            "Requirement already satisfied: flatbuffers>=2.0 in /usr/local/lib/python3.10/dist-packages (from mediapipe) (23.5.26)\n",
            "Requirement already satisfied: matplotlib in /usr/local/lib/python3.10/dist-packages (from mediapipe) (3.7.1)\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.10/dist-packages (from mediapipe) (1.23.5)\n",
            "Requirement already satisfied: opencv-contrib-python in /usr/local/lib/python3.10/dist-packages (from mediapipe) (4.8.0.76)\n",
            "Requirement already satisfied: protobuf<4,>=3.11 in /usr/local/lib/python3.10/dist-packages (from mediapipe) (3.20.3)\n",
            "Collecting sounddevice>=0.4.4 (from mediapipe)\n",
            "  Downloading sounddevice-0.4.6-py3-none-any.whl (31 kB)\n",
            "Requirement already satisfied: CFFI>=1.0 in /usr/local/lib/python3.10/dist-packages (from sounddevice>=0.4.4->mediapipe) (1.16.0)\n",
            "Requirement already satisfied: contourpy>=1.0.1 in /usr/local/lib/python3.10/dist-packages (from matplotlib->mediapipe) (1.2.0)\n",
            "Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.10/dist-packages (from matplotlib->mediapipe) (0.12.1)\n",
            "Requirement already satisfied: fonttools>=4.22.0 in /usr/local/lib/python3.10/dist-packages (from matplotlib->mediapipe) (4.46.0)\n",
            "Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.10/dist-packages (from matplotlib->mediapipe) (1.4.5)\n",
            "Requirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.10/dist-packages (from matplotlib->mediapipe) (23.2)\n",
            "Requirement already satisfied: pillow>=6.2.0 in /usr/local/lib/python3.10/dist-packages (from matplotlib->mediapipe) (9.4.0)\n",
            "Requirement already satisfied: pyparsing>=2.3.1 in /usr/local/lib/python3.10/dist-packages (from matplotlib->mediapipe) (3.1.1)\n",
            "Requirement already satisfied: python-dateutil>=2.7 in /usr/local/lib/python3.10/dist-packages (from matplotlib->mediapipe) (2.8.2)\n",
            "Requirement already satisfied: pycparser in /usr/local/lib/python3.10/dist-packages (from CFFI>=1.0->sounddevice>=0.4.4->mediapipe) (2.21)\n",
            "Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.10/dist-packages (from python-dateutil>=2.7->matplotlib->mediapipe) (1.16.0)\n",
            "Installing collected packages: sounddevice, mediapipe\n",
            "Successfully installed mediapipe-0.10.8 sounddevice-0.4.6\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "video_location = '/content/test.mp4'\n",
        "import os\n",
        "\n",
        "\n",
        "# Read images with OpenCV.\n",
        "#images= None\n",
        "image_dir = '/content/MEDIAPIPEinput/'\n",
        "os.makedirs(image_dir, exist_ok=True)\n",
        "image_dir_out = '/content/annotated_images'\n",
        "os.makedirs(image_dir_out, exist_ok=True)\n",
        "frame_rate = 24\n",
        "!ffmpeg -y \\\n",
        "  -i {video_location} \\\n",
        "  -r {frame_rate} {image_dir}out_%09d.png\n",
        "\n",
        "imgs_list = os.listdir(image_dir)\n",
        "imgs_list.sort()\n",
        "imgs_path = [os.path.join(image_dir, i) for i in imgs_list if i != 'outputs']\n"
      ],
      "metadata": {
        "id": "XXwQvA_2bMmf",
        "outputId": "4f8070db-7c8f-4503-c68f-c53ba459f2a3",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "ffmpeg version 4.4.2-0ubuntu0.22.04.1 Copyright (c) 2000-2021 the FFmpeg developers\n",
            "  built with gcc 11 (Ubuntu 11.2.0-19ubuntu1)\n",
            "  configuration: --prefix=/usr --extra-version=0ubuntu0.22.04.1 --toolchain=hardened --libdir=/usr/lib/x86_64-linux-gnu --incdir=/usr/include/x86_64-linux-gnu --arch=amd64 --enable-gpl --disable-stripping --enable-gnutls --enable-ladspa --enable-libaom --enable-libass --enable-libbluray --enable-libbs2b --enable-libcaca --enable-libcdio --enable-libcodec2 --enable-libdav1d --enable-libflite --enable-libfontconfig --enable-libfreetype --enable-libfribidi --enable-libgme --enable-libgsm --enable-libjack --enable-libmp3lame --enable-libmysofa --enable-libopenjpeg --enable-libopenmpt --enable-libopus --enable-libpulse --enable-librabbitmq --enable-librubberband --enable-libshine --enable-libsnappy --enable-libsoxr --enable-libspeex --enable-libsrt --enable-libssh --enable-libtheora --enable-libtwolame --enable-libvidstab --enable-libvorbis --enable-libvpx --enable-libwebp --enable-libx265 --enable-libxml2 --enable-libxvid --enable-libzimg --enable-libzmq --enable-libzvbi --enable-lv2 --enable-omx --enable-openal --enable-opencl --enable-opengl --enable-sdl2 --enable-pocketsphinx --enable-librsvg --enable-libmfx --enable-libdc1394 --enable-libdrm --enable-libiec61883 --enable-chromaprint --enable-frei0r --enable-libx264 --enable-shared\n",
            "  libavutil      56. 70.100 / 56. 70.100\n",
            "  libavcodec     58.134.100 / 58.134.100\n",
            "  libavformat    58. 76.100 / 58. 76.100\n",
            "  libavdevice    58. 13.100 / 58. 13.100\n",
            "  libavfilter     7.110.100 /  7.110.100\n",
            "  libswscale      5.  9.100 /  5.  9.100\n",
            "  libswresample   3.  9.100 /  3.  9.100\n",
            "  libpostproc    55.  9.100 / 55.  9.100\n",
            "Input #0, mov,mp4,m4a,3gp,3g2,mj2, from '/content/test.mp4':\n",
            "  Metadata:\n",
            "    major_brand     : mp42\n",
            "    minor_version   : 0\n",
            "    compatible_brands: mp42mp41iso4\n",
            "    creation_time   : 2021-12-09T21:02:50.000000Z\n",
            "  Duration: 00:00:40.58, start: 0.000000, bitrate: 673 kb/s\n",
            "  Stream #0:0(und): Video: h264 (Main) (avc1 / 0x31637661), yuv420p(tv, bt709), 1050x440, 601 kb/s, 25 fps, 25 tbr, 12800 tbn, 25600 tbc (default)\n",
            "    Metadata:\n",
            "      creation_time   : 2021-12-09T21:02:50.000000Z\n",
            "      handler_name    : Vireo Eyes v2.7.3\n",
            "      vendor_id       : [0][0][0][0]\n",
            "      encoder         : AVC Coding\n",
            "  Stream #0:1(und): Audio: aac (LC) (mp4a / 0x6134706D), 48000 Hz, stereo, fltp, 65 kb/s (default)\n",
            "    Metadata:\n",
            "      creation_time   : 2021-12-09T21:02:50.000000Z\n",
            "      handler_name    : Vireo Ears v2.7.3\n",
            "      vendor_id       : [0][0][0][0]\n",
            "Stream mapping:\n",
            "  Stream #0:0 -> #0:0 (h264 (native) -> png (native))\n",
            "Press [q] to stop, [?] for help\n",
            "Output #0, image2, to '/content/MEDIAPIPEinput/out_%09d.png':\n",
            "  Metadata:\n",
            "    major_brand     : mp42\n",
            "    minor_version   : 0\n",
            "    compatible_brands: mp42mp41iso4\n",
            "    encoder         : Lavf58.76.100\n",
            "  Stream #0:0(und): Video: png, rgb24(pc, gbr/bt709/bt709, progressive), 1050x440, q=2-31, 200 kb/s, 24 fps, 24 tbn (default)\n",
            "    Metadata:\n",
            "      creation_time   : 2021-12-09T21:02:50.000000Z\n",
            "      handler_name    : Vireo Eyes v2.7.3\n",
            "      vendor_id       : [0][0][0][0]\n",
            "      encoder         : Lavc58.134.100 png\n",
            "frame=  975 fps= 20 q=-0.0 Lsize=N/A time=00:00:40.62 bitrate=N/A dup=0 drop=39 speed=0.843x    \n",
            "video:665380kB audio:0kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: unknown\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import cv2\n",
        "from google.colab.patches import cv2_imshow\n",
        "import numpy as np\n",
        "import mediapipe as mp\n",
        "# mp_holistic = mp.solutions.holistic\n",
        "mp_pose = mp.solutions.pose\n",
        "!rm -r {image_dir}.ipynb_checkpoints\n",
        "\n",
        "# Run MediaPipe Pose with `enable_segmentation=True` to get pose segmentation.\n",
        "with mp_pose.Pose(static_image_mode=True,\n",
        "                          min_detection_confidence=0.2,\n",
        "                          model_complexity=2,\n",
        "                          enable_segmentation=True,) as pose:\n",
        "  temp_segmentation_mask =[]\n",
        "  for name, image in enumerate(imgs_path):\n",
        "    !rm -r {image_dir}.ipynb_checkpoints\n",
        "    # Convert the BGR image to RGB and process it with MediaPipe Pose.\n",
        "    image = cv2.imread(image)\n",
        "    results = pose.process(image)\n",
        "\n",
        "    # Draw pose segmentation.\n",
        "    print(f'Pose segmentation of {name}:')\n",
        "    annotated_image_pose = image.copy()\n",
        "    red_img = np.zeros_like(annotated_image_pose, dtype=np.uint8)\n",
        "    red_img[:, :] = (255,255,255)\n",
        "    ###check if segmentation_mask exists or not ## if exists then ok Else use previous mask temporarily\n",
        "    if results.segmentation_mask is None:\n",
        "      print(\"true\")\n",
        "      results.segmentation_mask = temp_segmentation_mask[-1]\n",
        "      temp_segmentation_mask.append(results.segmentation_mask)\n",
        "    else:\n",
        "      temp_segmentation_mask.append(results.segmentation_mask)\n",
        "    ###End check if segmentation_mask exists or not ## if exists then ok Else use previous mask temporarily\n",
        "    segm_2class = 0.0 + 1.0 * results.segmentation_mask\n",
        "    segm_2class = np.repeat(segm_2class[..., np.newaxis], 3, axis=2)\n",
        "    annotated_image_pose = annotated_image_pose * segm_2class + red_img * (1 - segm_2class)\n",
        "    #resize_and_show(annotated_image)\n",
        "    cv2.imwrite('%s/%s' %(image_dir_out, imgs_list[name]), annotated_image_pose)\n",
        "    !rm -r {image_dir_out}.ipynb_checkpoints\n",
        "\n",
        "# # Run MediaPipe Pose with `enable_segmentation=True` to get pose segmentation.\n",
        "# with mp_holistic.Holistic(static_image_mode=False,\n",
        "#                           min_detection_confidence=0.5,\n",
        "#                           model_complexity=2,\n",
        "#                           enable_segmentation=True,) as holistic:\n",
        "#   temp_segmentation_mask =[]\n",
        "#   for name, image in enumerate(imgs_path):\n",
        "#     !rm -r {image_dir}.ipynb_checkpoints\n",
        "#     # Convert the BGR image to RGB and process it with MediaPipe Pose.\n",
        "#     image = cv2.imread(image)\n",
        "#     results = holistic.process(image)\n",
        "\n",
        "#     # Draw pose segmentation.\n",
        "#     print(f'Pose segmentation of {name}:')\n",
        "#     annotated_image_pose = image.copy()\n",
        "#     red_img = np.zeros_like(annotated_image_pose, dtype=np.uint8)\n",
        "#     red_img[:, :] = (255,255,255)\n",
        "#     if results.segmentation_mask is None:\n",
        "#       print(\"true\")\n",
        "#       results.segmentation_mask = temp_segmentation_mask[-1]\n",
        "#       temp_segmentation_mask.append(results.segmentation_mask)\n",
        "#     else:\n",
        "#       temp_segmentation_mask.append(results.segmentation_mask)\n",
        "#     segm_2class = 0.2 + 0.8 * results.segmentation_mask\n",
        "#     segm_2class = np.repeat(segm_2class[..., np.newaxis], 3, axis=2)\n",
        "#     annotated_image_pose = annotated_image_pose * segm_2class + red_img * (1 - segm_2class)\n",
        "#     #resize_and_show(annotated_image)\n",
        "#     cv2.imwrite('%s/%s' %(image_dir_out, imgs_list[name]), annotated_image_pose)\n",
        "#     !rm -r {image_dir_out}.ipynb_checkpoints"
      ],
      "metadata": {
        "id": "SNuLCc3IYPsQ"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "video_location = '/content/test.mp4'\n",
        "image_dir = '/content/MEDIAPIPEinput/'\n",
        "image_dir_out = '/content/annotated_images'\n",
        "\n",
        "frame_rate = 24\n",
        "#create video 1\n",
        "!ffmpeg -y \\\n",
        "  -framerate {frame_rate} \\\n",
        "  -pattern_type glob \\\n",
        "  -i {image_dir}'/*.png' \\\n",
        "  inputA.mp4\n",
        "#create vodeo 2\n",
        "!ffmpeg \\\n",
        "  -framerate {frame_rate} \\\n",
        "  -pattern_type glob \\\n",
        "  -i {image_dir_out}'/*.png' \\\n",
        "  inputB.mp4\n",
        "#merge video 1 and video 2\n",
        "!ffmpeg \\\n",
        "  -i inputA.mp4 -i inputB.mp4 \\\n",
        "  -filter_complex \\\n",
        "  \"[0:v][1:v]vstack=inputs=2\" \\\n",
        "  finalOutput.mp4\n",
        "#add sound of main video to final output video\n",
        "!ffmpeg \\\n",
        "  -i 'finalOutput.mp4' -i {video_location} \\\n",
        "  -c:v copy -c:a copy -map 0:v:0 -map 1:a:0 \\\n",
        "  OUTPUT_FILE.mp4"
      ],
      "metadata": {
        "id": "-0ag81xGoP3_",
        "outputId": "13d0af2f-bf49-4707-dc0d-b7fcdd665d69",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 391
        }
      },
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "error",
          "ename": "NotImplementedError",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mNotImplementedError\u001b[0m                       Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-18-0fada02d2c9e>\u001b[0m in \u001b[0;36m<cell line: 7>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0mframe_rate\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m24\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      6\u001b[0m \u001b[0;31m#create video 1\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 7\u001b[0;31m \u001b[0mget_ipython\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msystem\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"ffmpeg -y    -framerate {frame_rate}    -pattern_type glob    -i {image_dir}'/*.png'    inputA.mp4\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      8\u001b[0m \u001b[0;31m#create vodeo 2\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      9\u001b[0m \u001b[0mget_ipython\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msystem\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"ffmpeg    -framerate {frame_rate}    -pattern_type glob    -i {image_dir_out}'/*.png'    inputB.mp4\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.10/dist-packages/google/colab/_shell.py\u001b[0m in \u001b[0;36msystem\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m     97\u001b[0m       \u001b[0mkwargs\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mupdate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m{\u001b[0m\u001b[0;34m'also_return_output'\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;32mTrue\u001b[0m\u001b[0;34m}\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     98\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 99\u001b[0;31m     \u001b[0moutput\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_system_commands\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_system_compat\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m  \u001b[0;31m# pylint:disable=protected-access\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    100\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    101\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mpip_warn\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.10/dist-packages/google/colab/_system_commands.py\u001b[0m in \u001b[0;36m_system_compat\u001b[0;34m(shell, cmd, also_return_output)\u001b[0m\n\u001b[1;32m    452\u001b[0m   \u001b[0;31m# is expected to call this function, thus adding one level of nesting to the\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    453\u001b[0m   \u001b[0;31m# stack.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 454\u001b[0;31m   result = _run_command(\n\u001b[0m\u001b[1;32m    455\u001b[0m       \u001b[0mshell\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvar_expand\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcmd\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdepth\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m2\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mclear_streamed_output\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    456\u001b[0m   )\n",
            "\u001b[0;32m/usr/local/lib/python3.10/dist-packages/google/colab/_system_commands.py\u001b[0m in \u001b[0;36m_run_command\u001b[0;34m(cmd, clear_streamed_output)\u001b[0m\n\u001b[1;32m    166\u001b[0m     \u001b[0mlocale_encoding\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mlocale\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mgetpreferredencoding\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    167\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mlocale_encoding\u001b[0m \u001b[0;34m!=\u001b[0m \u001b[0m_ENCODING\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 168\u001b[0;31m       raise NotImplementedError(\n\u001b[0m\u001b[1;32m    169\u001b[0m           \u001b[0;34m'A UTF-8 locale is required. Got {}'\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mformat\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlocale_encoding\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    170\u001b[0m       )\n",
            "\u001b[0;31mNotImplementedError\u001b[0m: A UTF-8 locale is required. Got ANSI_X3.4-1968"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# **[MediaPipe](https://github.com/google/mediapipe)** with ( **[EffficientSAM](https://arxiv.org/abs/2312.00863)** + **[Segment Anything](https://arxiv.org/abs/2304.02643)** )"
      ],
      "metadata": {
        "id": "1Jz1ICurvvwJ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "\n",
        "!wget -nc https://github.com/1kaiser/Media-Segment-Depth-MLP/releases/download/v0.2/efficientsam_s_gpu.jit\n",
        "!wget -nc https://github.com/1kaiser/Media-Segment-Depth-MLP/releases/download/v0.2/efficientsam_ti_gpu.jit\n"
      ],
      "metadata": {
        "id": "P_1IgIC-vhXI"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "\n",
        "import matplotlib.pyplot as plt\n",
        "import cv2\n",
        "import numpy as np\n",
        "import torch\n",
        "from torchvision.transforms import ToTensor\n",
        "from PIL import Image\n",
        "import io\n",
        "\n",
        "model = torch.jit.load('efficientsam_s_gpu.jit')\n",
        "\n",
        "def run_ours_point(img_path, pts_sampled, model):\n",
        "    image = cv2.imread(image_path)\n",
        "    image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)\n",
        "\n",
        "    img_tensor = ToTensor()(image)\n",
        "    pts_sampled = torch.reshape(torch.tensor(pts_sampled), [1, 1, -1, 2])\n",
        "    max_num_pts = pts_sampled.shape[2]\n",
        "    pts_labels = torch.ones(1, 1, max_num_pts)\n",
        "\n",
        "    predicted_logits, predicted_iou = model(img_tensor[None, ...].cuda(),pts_sampled.cuda(),pts_labels.cuda())\n",
        "    predicted_logits = predicted_logits.cpu()\n",
        "    all_masks = torch.ge(torch.sigmoid(predicted_logits[0, 0, :, :, :]), 0.5).numpy()\n",
        "    predicted_iou = predicted_iou[0, 0, ...].cpu().detach().numpy()\n",
        "\n",
        "    max_predicted_iou = -1\n",
        "    selected_mask_using_predicted_iou = None\n",
        "    for m in range(all_masks.shape[0]):\n",
        "        curr_predicted_iou = predicted_iou[m]\n",
        "        if (\n",
        "            curr_predicted_iou > max_predicted_iou\n",
        "            or selected_mask_using_predicted_iou is None\n",
        "        ):\n",
        "            max_predicted_iou = curr_predicted_iou\n",
        "            selected_mask_using_predicted_iou = all_masks[m]\n",
        "    return selected_mask_using_predicted_iou\n",
        "\n",
        "################################################################################\n",
        "import cv2\n",
        "import numpy as np\n",
        "import mediapipe as mp\n",
        "\n",
        "mp_pose = mp.solutions.pose\n",
        "pose = mp_pose.Pose(static_image_mode=True, min_detection_confidence=0.2, model_complexity=2, enable_segmentation=True)\n",
        "\n",
        "temp_segmentation_mask = []\n",
        "\n",
        "for idx, image_path in enumerate(imgs_path):\n",
        "    # Read and process the image.\n",
        "    image = cv2.imread(image_path)\n",
        "    results = pose.process(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))\n",
        "\n",
        "    # Check for segmentation mask existence.\n",
        "    if results.segmentation_mask is None:\n",
        "        if temp_segmentation_mask:\n",
        "            results.segmentation_mask = temp_segmentation_mask[-1]\n",
        "        else:\n",
        "            # Handle the case where the first image has no segmentation mask.\n",
        "            continue\n",
        "    temp_segmentation_mask.append(results.segmentation_mask)\n",
        "    _, binary_mask = cv2.threshold(results.segmentation_mask, 0.5, 255, cv2.THRESH_BINARY)\n",
        "    binary_mask = np.uint8(binary_mask)\n",
        "    contours, _ = cv2.findContours(binary_mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)\n",
        "    cx, cy = 0, 0\n",
        "    if contours:\n",
        "        # Assuming the largest contour corresponds to the object\n",
        "        object_contour = max(contours, key=cv2.contourArea)\n",
        "        # Calculate the centroid\n",
        "        M = cv2.moments(object_contour)\n",
        "        if M[\"m00\"] != 0:\n",
        "            cx = int(M[\"m10\"] / M[\"m00\"])\n",
        "            cy = int(M[\"m01\"] / M[\"m00\"])\n",
        "        else:\n",
        "            # Default centroid position in case of division by zero\n",
        "            cx, cy = 0, 0\n",
        "        print(f\"Centroid of the object in {image_path} is at: ({cx}, {cy})\")\n",
        "    else:\n",
        "        print(f\"No contours found in {image_path}\")\n",
        "\n",
        "    input_point = np.array([[cx, cy]])\n",
        "    mask = run_ours_point(image_path, input_point, model).astype(np.uint8)\n",
        "\n",
        "    # Create an annotated image with pose segmentation\n",
        "    segm_2class = np.repeat(mask[..., np.newaxis], 3, axis=2)\n",
        "    red_img = np.zeros_like(image, dtype=np.uint8) + (255, 255, 255)\n",
        "    annotated_image_pose = image * segm_2class + red_img * (1 - segm_2class)\n",
        "\n",
        "    # Save the annotated image\n",
        "    cv2.imwrite(f'{image_dir_out}/{imgs_list[idx]}', annotated_image_pose)\n",
        "\n",
        "# Close the pose object\n",
        "pose.close()\n"
      ],
      "metadata": {
        "id": "OYibbtkbsEAS"
      },
      "execution_count": 15,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "import subprocess\n",
        "\n",
        "frame_rate = 24\n",
        "image_dir = '/content/MEDIAPIPEinput/'\n",
        "image_dir_out = '/content/annotated_images'\n",
        "video_location = '/content/test.mp4'\n",
        "\n",
        "commands = [\n",
        "    f\"ffmpeg -y -framerate {frame_rate} -pattern_type glob -i '{image_dir}/*.png' inputA.mp4\",\n",
        "    f\"ffmpeg -framerate {frame_rate} -pattern_type glob -i '{image_dir_out}/*.png' inputB.mp4\",\n",
        "    \"ffmpeg -i inputA.mp4 -i inputB.mp4 -filter_complex '[0:v][1:v]vstack=inputs=2' finalOutput.mp4\",\n",
        "    f\"ffmpeg -i 'finalOutput.mp4' -i {video_location} -c:v copy -c:a copy -map 0:v:0 -map 1:a:0 OUTPUT_FILE.mp4\"\n",
        "]\n",
        "\n",
        "for cmd in commands:\n",
        "    subprocess.run(cmd, shell=True, check=True)\n"
      ],
      "metadata": {
        "id": "_hknPLUL2eEl"
      },
      "execution_count": 23,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "##***BoostingMonocularDepth code***"
      ],
      "metadata": {
        "id": "1G67vqC1-O3-"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Code section 1\n",
        "INPUT_FILES = 'input1'\n",
        "OUTPUT_FILES = 'outputs_leres'\n",
        "!mkdir -p {INPUT_FILES}\n",
        "!mkdir -p {OUTPUT_FILES}"
      ],
      "metadata": {
        "id": "HrdbNfdQcIDq"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "!ls"
      ],
      "metadata": {
        "id": "cUXERYUmqsyV"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "!ffmpeg \\\n",
        "  -i '/content/test.mp4' \\\n",
        "  -r 24 {INPUT_FILES}/out_%09d.png"
      ],
      "metadata": {
        "id": "ZBJJQKD4azSI",
        "outputId": "0dce7137-fe0f-456f-ee79-67d922f310c8",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "ffmpeg version 4.4.2-0ubuntu0.22.04.1 Copyright (c) 2000-2021 the FFmpeg developers\n",
            "  built with gcc 11 (Ubuntu 11.2.0-19ubuntu1)\n",
            "  configuration: --prefix=/usr --extra-version=0ubuntu0.22.04.1 --toolchain=hardened --libdir=/usr/lib/x86_64-linux-gnu --incdir=/usr/include/x86_64-linux-gnu --arch=amd64 --enable-gpl --disable-stripping --enable-gnutls --enable-ladspa --enable-libaom --enable-libass --enable-libbluray --enable-libbs2b --enable-libcaca --enable-libcdio --enable-libcodec2 --enable-libdav1d --enable-libflite --enable-libfontconfig --enable-libfreetype --enable-libfribidi --enable-libgme --enable-libgsm --enable-libjack --enable-libmp3lame --enable-libmysofa --enable-libopenjpeg --enable-libopenmpt --enable-libopus --enable-libpulse --enable-librabbitmq --enable-librubberband --enable-libshine --enable-libsnappy --enable-libsoxr --enable-libspeex --enable-libsrt --enable-libssh --enable-libtheora --enable-libtwolame --enable-libvidstab --enable-libvorbis --enable-libvpx --enable-libwebp --enable-libx265 --enable-libxml2 --enable-libxvid --enable-libzimg --enable-libzmq --enable-libzvbi --enable-lv2 --enable-omx --enable-openal --enable-opencl --enable-opengl --enable-sdl2 --enable-pocketsphinx --enable-librsvg --enable-libmfx --enable-libdc1394 --enable-libdrm --enable-libiec61883 --enable-chromaprint --enable-frei0r --enable-libx264 --enable-shared\n",
            "  libavutil      56. 70.100 / 56. 70.100\n",
            "  libavcodec     58.134.100 / 58.134.100\n",
            "  libavformat    58. 76.100 / 58. 76.100\n",
            "  libavdevice    58. 13.100 / 58. 13.100\n",
            "  libavfilter     7.110.100 /  7.110.100\n",
            "  libswscale      5.  9.100 /  5.  9.100\n",
            "  libswresample   3.  9.100 /  3.  9.100\n",
            "  libpostproc    55.  9.100 / 55.  9.100\n",
            "Input #0, mov,mp4,m4a,3gp,3g2,mj2, from '/content/test.mp4':\n",
            "  Metadata:\n",
            "    major_brand     : mp42\n",
            "    minor_version   : 0\n",
            "    compatible_brands: mp42mp41iso4\n",
            "    creation_time   : 2021-12-09T21:02:50.000000Z\n",
            "  Duration: 00:00:40.58, start: 0.000000, bitrate: 673 kb/s\n",
            "  Stream #0:0(und): Video: h264 (Main) (avc1 / 0x31637661), yuv420p(tv, bt709), 1050x440, 601 kb/s, 25 fps, 25 tbr, 12800 tbn, 25600 tbc (default)\n",
            "    Metadata:\n",
            "      creation_time   : 2021-12-09T21:02:50.000000Z\n",
            "      handler_name    : Vireo Eyes v2.7.3\n",
            "      vendor_id       : [0][0][0][0]\n",
            "      encoder         : AVC Coding\n",
            "  Stream #0:1(und): Audio: aac (LC) (mp4a / 0x6134706D), 48000 Hz, stereo, fltp, 65 kb/s (default)\n",
            "    Metadata:\n",
            "      creation_time   : 2021-12-09T21:02:50.000000Z\n",
            "      handler_name    : Vireo Ears v2.7.3\n",
            "      vendor_id       : [0][0][0][0]\n",
            "Stream mapping:\n",
            "  Stream #0:0 -> #0:0 (h264 (native) -> png (native))\n",
            "Press [q] to stop, [?] for help\n",
            "Output #0, image2, to 'input1/out_%09d.png':\n",
            "  Metadata:\n",
            "    major_brand     : mp42\n",
            "    minor_version   : 0\n",
            "    compatible_brands: mp42mp41iso4\n",
            "    encoder         : Lavf58.76.100\n",
            "  Stream #0:0(und): Video: png, rgb24(pc, gbr/bt709/bt709, progressive), 1050x440, q=2-31, 200 kb/s, 24 fps, 24 tbn (default)\n",
            "    Metadata:\n",
            "      creation_time   : 2021-12-09T21:02:50.000000Z\n",
            "      handler_name    : Vireo Eyes v2.7.3\n",
            "      vendor_id       : [0][0][0][0]\n",
            "      encoder         : Lavc58.134.100 png\n",
            "frame=  975 fps= 20 q=-0.0 Lsize=N/A time=00:00:40.62 bitrate=N/A dup=0 drop=39 speed=0.846x    \n",
            "video:665380kB audio:0kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: unknown\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "!wget -nc https://github.com/1kaiser/Media-Segment-Depth-MLP/releases/download/v0.2/latest_net_G.pth\n",
        "!wget -nc https://github.com/1kaiser/Media-Segment-Depth-MLP/releases/download/v0.2/res101.pth\n",
        "\n"
      ],
      "metadata": {
        "id": "UAdl2bf3kVol",
        "outputId": "27bdad53-1af4-417c-b709-b7d6ef644649",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "--2023-12-06 05:38:06--  https://github.com/1kaiser/Media-Segment-Depth-MLP/releases/download/v0.2/latest_net_G.pth\n",
            "Resolving github.com (github.com)... 140.82.113.3\n",
            "Connecting to github.com (github.com)|140.82.113.3|:443... connected.\n",
            "HTTP request sent, awaiting response... 302 Found\n",
            "Location: https://objects.githubusercontent.com/github-production-release-asset-2e65be/578313074/acdc51e6-0be4-40a7-9c03-aa4c3f4479a3?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=AKIAIWNJYAX4CSVEH53A%2F20231206%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20231206T053806Z&X-Amz-Expires=300&X-Amz-Signature=003557c0c329f6c8b5026fd0e4cc2e24190dc443269fa6a1d742976c954e6c66&X-Amz-SignedHeaders=host&actor_id=0&key_id=0&repo_id=578313074&response-content-disposition=attachment%3B%20filename%3Dlatest_net_G.pth&response-content-type=application%2Foctet-stream [following]\n",
            "--2023-12-06 05:38:06--  https://objects.githubusercontent.com/github-production-release-asset-2e65be/578313074/acdc51e6-0be4-40a7-9c03-aa4c3f4479a3?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=AKIAIWNJYAX4CSVEH53A%2F20231206%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20231206T053806Z&X-Amz-Expires=300&X-Amz-Signature=003557c0c329f6c8b5026fd0e4cc2e24190dc443269fa6a1d742976c954e6c66&X-Amz-SignedHeaders=host&actor_id=0&key_id=0&repo_id=578313074&response-content-disposition=attachment%3B%20filename%3Dlatest_net_G.pth&response-content-type=application%2Foctet-stream\n",
            "Resolving objects.githubusercontent.com (objects.githubusercontent.com)... 185.199.108.133, 185.199.109.133, 185.199.110.133, ...\n",
            "Connecting to objects.githubusercontent.com (objects.githubusercontent.com)|185.199.108.133|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 318268048 (304M) [application/octet-stream]\n",
            "Saving to: ‘latest_net_G.pth’\n",
            "\n",
            "latest_net_G.pth    100%[===================>] 303.52M   134MB/s    in 2.3s    \n",
            "\n",
            "2023-12-06 05:38:09 (134 MB/s) - ‘latest_net_G.pth’ saved [318268048/318268048]\n",
            "\n",
            "--2023-12-06 05:38:09--  https://github.com/1kaiser/Media-Segment-Depth-MLP/releases/download/v0.2/res101.pth\n",
            "Resolving github.com (github.com)... 140.82.113.3\n",
            "Connecting to github.com (github.com)|140.82.113.3|:443... connected.\n",
            "HTTP request sent, awaiting response... 302 Found\n",
            "Location: https://objects.githubusercontent.com/github-production-release-asset-2e65be/578313074/3df6ae06-9be8-43ba-a761-bc037b62e2e0?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=AKIAIWNJYAX4CSVEH53A%2F20231206%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20231206T053809Z&X-Amz-Expires=300&X-Amz-Signature=c102b622086288941ca15659e631d8bb34cb0534ddecb107f9ee5b1445f6734e&X-Amz-SignedHeaders=host&actor_id=0&key_id=0&repo_id=578313074&response-content-disposition=attachment%3B%20filename%3Dres101.pth&response-content-type=application%2Foctet-stream [following]\n",
            "--2023-12-06 05:38:09--  https://objects.githubusercontent.com/github-production-release-asset-2e65be/578313074/3df6ae06-9be8-43ba-a761-bc037b62e2e0?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=AKIAIWNJYAX4CSVEH53A%2F20231206%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20231206T053809Z&X-Amz-Expires=300&X-Amz-Signature=c102b622086288941ca15659e631d8bb34cb0534ddecb107f9ee5b1445f6734e&X-Amz-SignedHeaders=host&actor_id=0&key_id=0&repo_id=578313074&response-content-disposition=attachment%3B%20filename%3Dres101.pth&response-content-type=application%2Foctet-stream\n",
            "Resolving objects.githubusercontent.com (objects.githubusercontent.com)... 185.199.108.133, 185.199.109.133, 185.199.110.133, ...\n",
            "Connecting to objects.githubusercontent.com (objects.githubusercontent.com)|185.199.108.133|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 530760553 (506M) [application/octet-stream]\n",
            "Saving to: ‘res101.pth’\n",
            "\n",
            "res101.pth          100%[===================>] 506.17M   140MB/s    in 3.8s    \n",
            "\n",
            "2023-12-06 05:38:13 (134 MB/s) - ‘res101.pth’ saved [530760553/530760553]\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Code section 2\n",
        "\n",
        "\n",
        "# Clone git repo\n",
        "!git clone https://github.com/1kaiser/BoostingMonocularDepth.git\n",
        "!mkdir -p BoostingMonocularDepth/pix2pix/checkpoints/mergemodel/\n",
        "!mv /content/latest_net_G.pth BoostingMonocularDepth/pix2pix/checkpoints/mergemodel/\n",
        "!mv /content/res101.pth BoostingMonocularDepth/\n"
      ],
      "metadata": {
        "id": "cOCuEvGmkvBh",
        "outputId": "d3836778-858f-4fb0-ca26-6bf64a4c8009",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Cloning into 'BoostingMonocularDepth'...\n",
            "remote: Enumerating objects: 273, done.\u001b[K\n",
            "remote: Counting objects: 100% (3/3), done.\u001b[K\n",
            "remote: Compressing objects: 100% (3/3), done.\u001b[K\n",
            "remote: Total 273 (delta 0), reused 1 (delta 0), pack-reused 270\u001b[K\n",
            "Receiving objects: 100% (273/273), 52.58 MiB | 38.35 MiB/s, done.\n",
            "Resolving deltas: 100% (98/98), done.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# # Code section 2\n",
        "\n",
        "\n",
        "# # Clone git repo\n",
        "# !git clone https://github.com/1kaiser/BoostingMonocularDepth.git\n",
        "\n",
        "# # Downloading merge model weights\n",
        "# !mkdir -p /content/BoostingMonocularDepth/pix2pix/checkpoints/mergemodel/\n",
        "# !gdown 1BbX2_NjrmrzRIB2CRJ-1rNRAjPABXPCh\n",
        "# !mv latest_net_G.pth /content/BoostingMonocularDepth/pix2pix/checkpoints/mergemodel/\n",
        "\n",
        "# # Downloading Midas weights\n",
        "# #!gdown https://drive.google.com/uc?id=1nqW_Hwj86kslfsXR7EnXpEWdO2csz1cC\n",
        "# #!mv model.pt /content/BoostingMonocularDepth/midas/\n",
        "\n",
        "# # # Downloading LeRes weights\n",
        "# !gdown 1vni3Oh4y0S4eatz_u_aDnwDLwxBjUpCh\n",
        "# !mv res101.pth /content/BoostingMonocularDepth/"
      ],
      "metadata": {
        "id": "MHjb_2jxbHjP"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Code section 3\n",
        "%cd BoostingMonocularDepth/"
      ],
      "metadata": {
        "id": "rAOyd3jnbfRu",
        "outputId": "2e163080-262a-4315-bb2d-43f6389dfa04",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "/content/BoostingMonocularDepth\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "check and delete  fiiles"
      ],
      "metadata": {
        "id": "QCNFJkXaR-ZA"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "INPUT_FILES = r'/content/input1'\n",
        "OUTPUT_FILES = r'/content/outputs_leres'"
      ],
      "metadata": {
        "id": "aF3FVhNbSPNJ"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "%%writefile goDeleteHolistic.js\n",
        "'use strict';\n",
        "\n",
        "const fs = require('fs');\n",
        "\n",
        "// Define the paths to your input and output directories\n",
        "const INPUT_FILES = '/content/input1';\n",
        "const OUTPUT_FILES = '/content/outputs_leres';\n",
        "\n",
        "// Read the contents of both directories\n",
        "let darray1 = fs.readdirSync(INPUT_FILES);\n",
        "let darray2 = fs.readdirSync(OUTPUT_FILES);\n",
        "let d = 0;\n",
        "\n",
        "// Iterate through the input directory files\n",
        "for (let index = 0; index < darray1.length; index++) {\n",
        "    // Log the current index\n",
        "    console.log(index);\n",
        "\n",
        "    // Check if the file from the input directory exists in the output directory\n",
        "    if (darray2.includes(darray1[index])) {\n",
        "        // Log the count of deleted files\n",
        "        console.log(d++);\n",
        "\n",
        "        // Construct the file path and delete the file\n",
        "        fs.unlinkSync(`${INPUT_FILES}/${darray1[index]}`);\n",
        "    }\n",
        "}\n",
        "\n"
      ],
      "metadata": {
        "id": "xoDYwaPER849",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "a4be8e4f-d987-4e49-df8f-3c6652e42d8c"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Overwriting goDeleteHolistic.js\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "!node goDeleteHolistic.js"
      ],
      "metadata": {
        "id": "tFtDRDGcSDMS",
        "outputId": "6f5919d0-71e5-470f-bfd9-8952a05a3df9",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[33m0\u001b[39m\n",
            "\u001b[33m1\u001b[39m\n",
            "\u001b[33m2\u001b[39m\n",
            "\u001b[33m3\u001b[39m\n",
            "\u001b[33m4\u001b[39m\n",
            "\u001b[33m5\u001b[39m\n",
            "\u001b[33m6\u001b[39m\n",
            "\u001b[33m7\u001b[39m\n",
            "\u001b[33m8\u001b[39m\n",
            "\u001b[33m9\u001b[39m\n",
            "\u001b[33m10\u001b[39m\n",
            "\u001b[33m11\u001b[39m\n",
            "\u001b[33m12\u001b[39m\n",
            "\u001b[33m13\u001b[39m\n",
            "\u001b[33m14\u001b[39m\n",
            "\u001b[33m15\u001b[39m\n",
            "\u001b[33m16\u001b[39m\n",
            "\u001b[33m17\u001b[39m\n",
            "\u001b[33m18\u001b[39m\n",
            "\u001b[33m19\u001b[39m\n",
            "\u001b[33m20\u001b[39m\n",
            "\u001b[33m21\u001b[39m\n",
            "\u001b[33m22\u001b[39m\n",
            "\u001b[33m23\u001b[39m\n",
            "\u001b[33m24\u001b[39m\n",
            "\u001b[33m25\u001b[39m\n",
            "\u001b[33m26\u001b[39m\n",
            "\u001b[33m27\u001b[39m\n",
            "\u001b[33m28\u001b[39m\n",
            "\u001b[33m29\u001b[39m\n",
            "\u001b[33m30\u001b[39m\n",
            "\u001b[33m31\u001b[39m\n",
            "\u001b[33m32\u001b[39m\n",
            "\u001b[33m33\u001b[39m\n",
            "\u001b[33m34\u001b[39m\n",
            "\u001b[33m35\u001b[39m\n",
            "\u001b[33m36\u001b[39m\n",
            "\u001b[33m37\u001b[39m\n",
            "\u001b[33m38\u001b[39m\n",
            "\u001b[33m39\u001b[39m\n",
            "\u001b[33m40\u001b[39m\n",
            "\u001b[33m41\u001b[39m\n",
            "\u001b[33m42\u001b[39m\n",
            "\u001b[33m43\u001b[39m\n",
            "\u001b[33m44\u001b[39m\n",
            "\u001b[33m45\u001b[39m\n",
            "\u001b[33m46\u001b[39m\n",
            "\u001b[33m47\u001b[39m\n",
            "\u001b[33m48\u001b[39m\n",
            "\u001b[33m49\u001b[39m\n",
            "\u001b[33m50\u001b[39m\n",
            "\u001b[33m51\u001b[39m\n",
            "\u001b[33m52\u001b[39m\n",
            "\u001b[33m53\u001b[39m\n",
            "\u001b[33m54\u001b[39m\n",
            "\u001b[33m55\u001b[39m\n",
            "\u001b[33m56\u001b[39m\n",
            "\u001b[33m57\u001b[39m\n",
            "\u001b[33m58\u001b[39m\n",
            "\u001b[33m59\u001b[39m\n",
            "\u001b[33m60\u001b[39m\n",
            "\u001b[33m61\u001b[39m\n",
            "\u001b[33m62\u001b[39m\n",
            "\u001b[33m63\u001b[39m\n",
            "\u001b[33m64\u001b[39m\n",
            "\u001b[33m65\u001b[39m\n",
            "\u001b[33m66\u001b[39m\n",
            "\u001b[33m67\u001b[39m\n",
            "\u001b[33m68\u001b[39m\n",
            "\u001b[33m69\u001b[39m\n",
            "\u001b[33m70\u001b[39m\n",
            "\u001b[33m71\u001b[39m\n",
            "\u001b[33m72\u001b[39m\n",
            "\u001b[33m73\u001b[39m\n",
            "\u001b[33m74\u001b[39m\n",
            "\u001b[33m75\u001b[39m\n",
            "\u001b[33m76\u001b[39m\n",
            "\u001b[33m77\u001b[39m\n",
            "\u001b[33m78\u001b[39m\n",
            "\u001b[33m79\u001b[39m\n",
            "\u001b[33m80\u001b[39m\n",
            "\u001b[33m81\u001b[39m\n",
            "\u001b[33m82\u001b[39m\n",
            "\u001b[33m83\u001b[39m\n",
            "\u001b[33m84\u001b[39m\n",
            "\u001b[33m85\u001b[39m\n",
            "\u001b[33m86\u001b[39m\n",
            "\u001b[33m87\u001b[39m\n",
            "\u001b[33m88\u001b[39m\n",
            "\u001b[33m89\u001b[39m\n",
            "\u001b[33m90\u001b[39m\n",
            "\u001b[33m91\u001b[39m\n",
            "\u001b[33m92\u001b[39m\n",
            "\u001b[33m93\u001b[39m\n",
            "\u001b[33m94\u001b[39m\n",
            "\u001b[33m95\u001b[39m\n",
            "\u001b[33m96\u001b[39m\n",
            "\u001b[33m97\u001b[39m\n",
            "\u001b[33m98\u001b[39m\n",
            "\u001b[33m99\u001b[39m\n",
            "\u001b[33m100\u001b[39m\n",
            "\u001b[33m101\u001b[39m\n",
            "\u001b[33m102\u001b[39m\n",
            "\u001b[33m103\u001b[39m\n",
            "\u001b[33m104\u001b[39m\n",
            "\u001b[33m105\u001b[39m\n",
            "\u001b[33m106\u001b[39m\n",
            "\u001b[33m107\u001b[39m\n",
            "\u001b[33m108\u001b[39m\n",
            "\u001b[33m109\u001b[39m\n",
            "\u001b[33m110\u001b[39m\n",
            "\u001b[33m111\u001b[39m\n",
            "\u001b[33m112\u001b[39m\n",
            "\u001b[33m113\u001b[39m\n",
            "\u001b[33m114\u001b[39m\n",
            "\u001b[33m115\u001b[39m\n",
            "\u001b[33m116\u001b[39m\n",
            "\u001b[33m117\u001b[39m\n",
            "\u001b[33m118\u001b[39m\n",
            "\u001b[33m119\u001b[39m\n",
            "\u001b[33m120\u001b[39m\n",
            "\u001b[33m121\u001b[39m\n",
            "\u001b[33m122\u001b[39m\n",
            "\u001b[33m123\u001b[39m\n",
            "\u001b[33m124\u001b[39m\n",
            "\u001b[33m125\u001b[39m\n",
            "\u001b[33m126\u001b[39m\n",
            "\u001b[33m127\u001b[39m\n",
            "\u001b[33m128\u001b[39m\n",
            "\u001b[33m129\u001b[39m\n",
            "\u001b[33m130\u001b[39m\n",
            "\u001b[33m131\u001b[39m\n",
            "\u001b[33m132\u001b[39m\n",
            "\u001b[33m133\u001b[39m\n",
            "\u001b[33m134\u001b[39m\n",
            "\u001b[33m135\u001b[39m\n",
            "\u001b[33m136\u001b[39m\n",
            "\u001b[33m137\u001b[39m\n",
            "\u001b[33m138\u001b[39m\n",
            "\u001b[33m139\u001b[39m\n",
            "\u001b[33m140\u001b[39m\n",
            "\u001b[33m141\u001b[39m\n",
            "\u001b[33m142\u001b[39m\n",
            "\u001b[33m143\u001b[39m\n",
            "\u001b[33m144\u001b[39m\n",
            "\u001b[33m145\u001b[39m\n",
            "\u001b[33m146\u001b[39m\n",
            "\u001b[33m147\u001b[39m\n",
            "\u001b[33m148\u001b[39m\n",
            "\u001b[33m149\u001b[39m\n",
            "\u001b[33m150\u001b[39m\n",
            "\u001b[33m151\u001b[39m\n",
            "\u001b[33m152\u001b[39m\n",
            "\u001b[33m153\u001b[39m\n",
            "\u001b[33m154\u001b[39m\n",
            "\u001b[33m155\u001b[39m\n",
            "\u001b[33m156\u001b[39m\n",
            "\u001b[33m157\u001b[39m\n",
            "\u001b[33m158\u001b[39m\n",
            "\u001b[33m159\u001b[39m\n",
            "\u001b[33m160\u001b[39m\n",
            "\u001b[33m161\u001b[39m\n",
            "\u001b[33m162\u001b[39m\n",
            "\u001b[33m163\u001b[39m\n",
            "\u001b[33m164\u001b[39m\n",
            "\u001b[33m165\u001b[39m\n",
            "\u001b[33m166\u001b[39m\n",
            "\u001b[33m167\u001b[39m\n",
            "\u001b[33m168\u001b[39m\n",
            "\u001b[33m169\u001b[39m\n",
            "\u001b[33m170\u001b[39m\n",
            "\u001b[33m171\u001b[39m\n",
            "\u001b[33m172\u001b[39m\n",
            "\u001b[33m173\u001b[39m\n",
            "\u001b[33m174\u001b[39m\n",
            "\u001b[33m175\u001b[39m\n",
            "\u001b[33m176\u001b[39m\n",
            "\u001b[33m177\u001b[39m\n",
            "\u001b[33m178\u001b[39m\n",
            "\u001b[33m179\u001b[39m\n",
            "\u001b[33m180\u001b[39m\n",
            "\u001b[33m181\u001b[39m\n",
            "\u001b[33m182\u001b[39m\n",
            "\u001b[33m183\u001b[39m\n",
            "\u001b[33m184\u001b[39m\n",
            "\u001b[33m185\u001b[39m\n",
            "\u001b[33m186\u001b[39m\n",
            "\u001b[33m187\u001b[39m\n",
            "\u001b[33m188\u001b[39m\n",
            "\u001b[33m189\u001b[39m\n",
            "\u001b[33m190\u001b[39m\n",
            "\u001b[33m191\u001b[39m\n",
            "\u001b[33m192\u001b[39m\n",
            "\u001b[33m193\u001b[39m\n",
            "\u001b[33m194\u001b[39m\n",
            "\u001b[33m195\u001b[39m\n",
            "\u001b[33m196\u001b[39m\n",
            "\u001b[33m197\u001b[39m\n",
            "\u001b[33m198\u001b[39m\n",
            "\u001b[33m199\u001b[39m\n",
            "\u001b[33m200\u001b[39m\n",
            "\u001b[33m201\u001b[39m\n",
            "\u001b[33m202\u001b[39m\n",
            "\u001b[33m203\u001b[39m\n",
            "\u001b[33m204\u001b[39m\n",
            "\u001b[33m205\u001b[39m\n",
            "\u001b[33m206\u001b[39m\n",
            "\u001b[33m207\u001b[39m\n",
            "\u001b[33m208\u001b[39m\n",
            "\u001b[33m209\u001b[39m\n",
            "\u001b[33m210\u001b[39m\n",
            "\u001b[33m211\u001b[39m\n",
            "\u001b[33m212\u001b[39m\n",
            "\u001b[33m213\u001b[39m\n",
            "\u001b[33m214\u001b[39m\n",
            "\u001b[33m215\u001b[39m\n",
            "\u001b[33m216\u001b[39m\n",
            "\u001b[33m217\u001b[39m\n",
            "\u001b[33m218\u001b[39m\n",
            "\u001b[33m219\u001b[39m\n",
            "\u001b[33m220\u001b[39m\n",
            "\u001b[33m221\u001b[39m\n",
            "\u001b[33m222\u001b[39m\n",
            "\u001b[33m223\u001b[39m\n",
            "\u001b[33m224\u001b[39m\n",
            "\u001b[33m225\u001b[39m\n",
            "\u001b[33m226\u001b[39m\n",
            "\u001b[33m227\u001b[39m\n",
            "\u001b[33m228\u001b[39m\n",
            "\u001b[33m229\u001b[39m\n",
            "\u001b[33m230\u001b[39m\n",
            "\u001b[33m231\u001b[39m\n",
            "\u001b[33m232\u001b[39m\n",
            "\u001b[33m233\u001b[39m\n",
            "\u001b[33m234\u001b[39m\n",
            "\u001b[33m235\u001b[39m\n",
            "\u001b[33m236\u001b[39m\n",
            "\u001b[33m237\u001b[39m\n",
            "\u001b[33m238\u001b[39m\n",
            "\u001b[33m239\u001b[39m\n",
            "\u001b[33m240\u001b[39m\n",
            "\u001b[33m241\u001b[39m\n",
            "\u001b[33m242\u001b[39m\n",
            "\u001b[33m243\u001b[39m\n",
            "\u001b[33m244\u001b[39m\n",
            "\u001b[33m245\u001b[39m\n",
            "\u001b[33m246\u001b[39m\n",
            "\u001b[33m247\u001b[39m\n",
            "\u001b[33m248\u001b[39m\n",
            "\u001b[33m249\u001b[39m\n",
            "\u001b[33m250\u001b[39m\n",
            "\u001b[33m251\u001b[39m\n",
            "\u001b[33m252\u001b[39m\n",
            "\u001b[33m253\u001b[39m\n",
            "\u001b[33m254\u001b[39m\n",
            "\u001b[33m255\u001b[39m\n",
            "\u001b[33m256\u001b[39m\n",
            "\u001b[33m257\u001b[39m\n",
            "\u001b[33m258\u001b[39m\n",
            "\u001b[33m259\u001b[39m\n",
            "\u001b[33m260\u001b[39m\n",
            "\u001b[33m261\u001b[39m\n",
            "\u001b[33m262\u001b[39m\n",
            "\u001b[33m263\u001b[39m\n",
            "\u001b[33m264\u001b[39m\n",
            "\u001b[33m265\u001b[39m\n",
            "\u001b[33m266\u001b[39m\n",
            "\u001b[33m267\u001b[39m\n",
            "\u001b[33m268\u001b[39m\n",
            "\u001b[33m269\u001b[39m\n",
            "\u001b[33m270\u001b[39m\n",
            "\u001b[33m271\u001b[39m\n",
            "\u001b[33m272\u001b[39m\n",
            "\u001b[33m273\u001b[39m\n",
            "\u001b[33m274\u001b[39m\n",
            "\u001b[33m275\u001b[39m\n",
            "\u001b[33m276\u001b[39m\n",
            "\u001b[33m277\u001b[39m\n",
            "\u001b[33m278\u001b[39m\n",
            "\u001b[33m279\u001b[39m\n",
            "\u001b[33m280\u001b[39m\n",
            "\u001b[33m281\u001b[39m\n",
            "\u001b[33m282\u001b[39m\n",
            "\u001b[33m283\u001b[39m\n",
            "\u001b[33m284\u001b[39m\n",
            "\u001b[33m285\u001b[39m\n",
            "\u001b[33m286\u001b[39m\n",
            "\u001b[33m287\u001b[39m\n",
            "\u001b[33m288\u001b[39m\n",
            "\u001b[33m289\u001b[39m\n",
            "\u001b[33m290\u001b[39m\n",
            "\u001b[33m291\u001b[39m\n",
            "\u001b[33m292\u001b[39m\n",
            "\u001b[33m293\u001b[39m\n",
            "\u001b[33m294\u001b[39m\n",
            "\u001b[33m295\u001b[39m\n",
            "\u001b[33m296\u001b[39m\n",
            "\u001b[33m297\u001b[39m\n",
            "\u001b[33m298\u001b[39m\n",
            "\u001b[33m299\u001b[39m\n",
            "\u001b[33m300\u001b[39m\n",
            "\u001b[33m301\u001b[39m\n",
            "\u001b[33m302\u001b[39m\n",
            "\u001b[33m303\u001b[39m\n",
            "\u001b[33m304\u001b[39m\n",
            "\u001b[33m305\u001b[39m\n",
            "\u001b[33m306\u001b[39m\n",
            "\u001b[33m307\u001b[39m\n",
            "\u001b[33m308\u001b[39m\n",
            "\u001b[33m309\u001b[39m\n",
            "\u001b[33m310\u001b[39m\n",
            "\u001b[33m311\u001b[39m\n",
            "\u001b[33m312\u001b[39m\n",
            "\u001b[33m313\u001b[39m\n",
            "\u001b[33m314\u001b[39m\n",
            "\u001b[33m315\u001b[39m\n",
            "\u001b[33m316\u001b[39m\n",
            "\u001b[33m317\u001b[39m\n",
            "\u001b[33m318\u001b[39m\n",
            "\u001b[33m319\u001b[39m\n",
            "\u001b[33m320\u001b[39m\n",
            "\u001b[33m321\u001b[39m\n",
            "\u001b[33m322\u001b[39m\n",
            "\u001b[33m323\u001b[39m\n",
            "\u001b[33m324\u001b[39m\n",
            "\u001b[33m325\u001b[39m\n",
            "\u001b[33m326\u001b[39m\n",
            "\u001b[33m327\u001b[39m\n",
            "\u001b[33m328\u001b[39m\n",
            "\u001b[33m329\u001b[39m\n",
            "\u001b[33m330\u001b[39m\n",
            "\u001b[33m331\u001b[39m\n",
            "\u001b[33m332\u001b[39m\n",
            "\u001b[33m333\u001b[39m\n",
            "\u001b[33m334\u001b[39m\n",
            "\u001b[33m335\u001b[39m\n",
            "\u001b[33m336\u001b[39m\n",
            "\u001b[33m337\u001b[39m\n",
            "\u001b[33m338\u001b[39m\n",
            "\u001b[33m339\u001b[39m\n",
            "\u001b[33m340\u001b[39m\n",
            "\u001b[33m341\u001b[39m\n",
            "\u001b[33m342\u001b[39m\n",
            "\u001b[33m343\u001b[39m\n",
            "\u001b[33m344\u001b[39m\n",
            "\u001b[33m345\u001b[39m\n",
            "\u001b[33m346\u001b[39m\n",
            "\u001b[33m347\u001b[39m\n",
            "\u001b[33m348\u001b[39m\n",
            "\u001b[33m349\u001b[39m\n",
            "\u001b[33m350\u001b[39m\n",
            "\u001b[33m351\u001b[39m\n",
            "\u001b[33m352\u001b[39m\n",
            "\u001b[33m353\u001b[39m\n",
            "\u001b[33m354\u001b[39m\n",
            "\u001b[33m355\u001b[39m\n",
            "\u001b[33m356\u001b[39m\n",
            "\u001b[33m357\u001b[39m\n",
            "\u001b[33m358\u001b[39m\n",
            "\u001b[33m359\u001b[39m\n",
            "\u001b[33m360\u001b[39m\n",
            "\u001b[33m361\u001b[39m\n",
            "\u001b[33m362\u001b[39m\n",
            "\u001b[33m363\u001b[39m\n",
            "\u001b[33m364\u001b[39m\n",
            "\u001b[33m365\u001b[39m\n",
            "\u001b[33m366\u001b[39m\n",
            "\u001b[33m367\u001b[39m\n",
            "\u001b[33m368\u001b[39m\n",
            "\u001b[33m369\u001b[39m\n",
            "\u001b[33m370\u001b[39m\n",
            "\u001b[33m371\u001b[39m\n",
            "\u001b[33m372\u001b[39m\n",
            "\u001b[33m373\u001b[39m\n",
            "\u001b[33m374\u001b[39m\n",
            "\u001b[33m375\u001b[39m\n",
            "\u001b[33m376\u001b[39m\n",
            "\u001b[33m377\u001b[39m\n",
            "\u001b[33m378\u001b[39m\n",
            "\u001b[33m379\u001b[39m\n",
            "\u001b[33m380\u001b[39m\n",
            "\u001b[33m381\u001b[39m\n",
            "\u001b[33m382\u001b[39m\n",
            "\u001b[33m383\u001b[39m\n",
            "\u001b[33m384\u001b[39m\n",
            "\u001b[33m385\u001b[39m\n",
            "\u001b[33m386\u001b[39m\n",
            "\u001b[33m387\u001b[39m\n",
            "\u001b[33m388\u001b[39m\n",
            "\u001b[33m389\u001b[39m\n",
            "\u001b[33m390\u001b[39m\n",
            "\u001b[33m391\u001b[39m\n",
            "\u001b[33m392\u001b[39m\n",
            "\u001b[33m393\u001b[39m\n",
            "\u001b[33m394\u001b[39m\n",
            "\u001b[33m395\u001b[39m\n",
            "\u001b[33m396\u001b[39m\n",
            "\u001b[33m397\u001b[39m\n",
            "\u001b[33m398\u001b[39m\n",
            "\u001b[33m399\u001b[39m\n",
            "\u001b[33m400\u001b[39m\n",
            "\u001b[33m401\u001b[39m\n",
            "\u001b[33m402\u001b[39m\n",
            "\u001b[33m403\u001b[39m\n",
            "\u001b[33m404\u001b[39m\n",
            "\u001b[33m405\u001b[39m\n",
            "\u001b[33m406\u001b[39m\n",
            "\u001b[33m407\u001b[39m\n",
            "\u001b[33m408\u001b[39m\n",
            "\u001b[33m409\u001b[39m\n",
            "\u001b[33m410\u001b[39m\n",
            "\u001b[33m411\u001b[39m\n",
            "\u001b[33m412\u001b[39m\n",
            "\u001b[33m413\u001b[39m\n",
            "\u001b[33m414\u001b[39m\n",
            "\u001b[33m415\u001b[39m\n",
            "\u001b[33m416\u001b[39m\n",
            "\u001b[33m417\u001b[39m\n",
            "\u001b[33m418\u001b[39m\n",
            "\u001b[33m419\u001b[39m\n",
            "\u001b[33m420\u001b[39m\n",
            "\u001b[33m421\u001b[39m\n",
            "\u001b[33m422\u001b[39m\n",
            "\u001b[33m423\u001b[39m\n",
            "\u001b[33m424\u001b[39m\n",
            "\u001b[33m425\u001b[39m\n",
            "\u001b[33m426\u001b[39m\n",
            "\u001b[33m427\u001b[39m\n",
            "\u001b[33m428\u001b[39m\n",
            "\u001b[33m429\u001b[39m\n",
            "\u001b[33m430\u001b[39m\n",
            "\u001b[33m431\u001b[39m\n",
            "\u001b[33m432\u001b[39m\n",
            "\u001b[33m433\u001b[39m\n",
            "\u001b[33m434\u001b[39m\n",
            "\u001b[33m435\u001b[39m\n",
            "\u001b[33m436\u001b[39m\n",
            "\u001b[33m437\u001b[39m\n",
            "\u001b[33m438\u001b[39m\n",
            "\u001b[33m439\u001b[39m\n",
            "\u001b[33m440\u001b[39m\n",
            "\u001b[33m441\u001b[39m\n",
            "\u001b[33m442\u001b[39m\n",
            "\u001b[33m443\u001b[39m\n",
            "\u001b[33m444\u001b[39m\n",
            "\u001b[33m445\u001b[39m\n",
            "\u001b[33m446\u001b[39m\n",
            "\u001b[33m447\u001b[39m\n",
            "\u001b[33m448\u001b[39m\n",
            "\u001b[33m449\u001b[39m\n",
            "\u001b[33m450\u001b[39m\n",
            "\u001b[33m451\u001b[39m\n",
            "\u001b[33m452\u001b[39m\n",
            "\u001b[33m453\u001b[39m\n",
            "\u001b[33m454\u001b[39m\n",
            "\u001b[33m455\u001b[39m\n",
            "\u001b[33m456\u001b[39m\n",
            "\u001b[33m457\u001b[39m\n",
            "\u001b[33m458\u001b[39m\n",
            "\u001b[33m459\u001b[39m\n",
            "\u001b[33m460\u001b[39m\n",
            "\u001b[33m461\u001b[39m\n",
            "\u001b[33m462\u001b[39m\n",
            "\u001b[33m463\u001b[39m\n",
            "\u001b[33m464\u001b[39m\n",
            "\u001b[33m465\u001b[39m\n",
            "\u001b[33m466\u001b[39m\n",
            "\u001b[33m467\u001b[39m\n",
            "\u001b[33m468\u001b[39m\n",
            "\u001b[33m469\u001b[39m\n",
            "\u001b[33m470\u001b[39m\n",
            "\u001b[33m471\u001b[39m\n",
            "\u001b[33m472\u001b[39m\n",
            "\u001b[33m473\u001b[39m\n",
            "\u001b[33m474\u001b[39m\n",
            "\u001b[33m475\u001b[39m\n",
            "\u001b[33m476\u001b[39m\n",
            "\u001b[33m477\u001b[39m\n",
            "\u001b[33m478\u001b[39m\n",
            "\u001b[33m479\u001b[39m\n",
            "\u001b[33m480\u001b[39m\n",
            "\u001b[33m481\u001b[39m\n",
            "\u001b[33m482\u001b[39m\n",
            "\u001b[33m483\u001b[39m\n",
            "\u001b[33m484\u001b[39m\n",
            "\u001b[33m485\u001b[39m\n",
            "\u001b[33m486\u001b[39m\n",
            "\u001b[33m487\u001b[39m\n",
            "\u001b[33m488\u001b[39m\n",
            "\u001b[33m489\u001b[39m\n",
            "\u001b[33m490\u001b[39m\n",
            "\u001b[33m491\u001b[39m\n",
            "\u001b[33m492\u001b[39m\n",
            "\u001b[33m493\u001b[39m\n",
            "\u001b[33m494\u001b[39m\n",
            "\u001b[33m495\u001b[39m\n",
            "\u001b[33m496\u001b[39m\n",
            "\u001b[33m497\u001b[39m\n",
            "\u001b[33m498\u001b[39m\n",
            "\u001b[33m499\u001b[39m\n",
            "\u001b[33m500\u001b[39m\n",
            "\u001b[33m501\u001b[39m\n",
            "\u001b[33m502\u001b[39m\n",
            "\u001b[33m503\u001b[39m\n",
            "\u001b[33m504\u001b[39m\n",
            "\u001b[33m505\u001b[39m\n",
            "\u001b[33m506\u001b[39m\n",
            "\u001b[33m507\u001b[39m\n",
            "\u001b[33m508\u001b[39m\n",
            "\u001b[33m509\u001b[39m\n",
            "\u001b[33m510\u001b[39m\n",
            "\u001b[33m511\u001b[39m\n",
            "\u001b[33m512\u001b[39m\n",
            "\u001b[33m513\u001b[39m\n",
            "\u001b[33m514\u001b[39m\n",
            "\u001b[33m515\u001b[39m\n",
            "\u001b[33m516\u001b[39m\n",
            "\u001b[33m517\u001b[39m\n",
            "\u001b[33m518\u001b[39m\n",
            "\u001b[33m519\u001b[39m\n",
            "\u001b[33m520\u001b[39m\n",
            "\u001b[33m521\u001b[39m\n",
            "\u001b[33m522\u001b[39m\n",
            "\u001b[33m523\u001b[39m\n",
            "\u001b[33m524\u001b[39m\n",
            "\u001b[33m525\u001b[39m\n",
            "\u001b[33m526\u001b[39m\n",
            "\u001b[33m527\u001b[39m\n",
            "\u001b[33m528\u001b[39m\n",
            "\u001b[33m529\u001b[39m\n",
            "\u001b[33m530\u001b[39m\n",
            "\u001b[33m531\u001b[39m\n",
            "\u001b[33m532\u001b[39m\n",
            "\u001b[33m533\u001b[39m\n",
            "\u001b[33m534\u001b[39m\n",
            "\u001b[33m535\u001b[39m\n",
            "\u001b[33m536\u001b[39m\n",
            "\u001b[33m537\u001b[39m\n",
            "\u001b[33m538\u001b[39m\n",
            "\u001b[33m539\u001b[39m\n",
            "\u001b[33m540\u001b[39m\n",
            "\u001b[33m541\u001b[39m\n",
            "\u001b[33m542\u001b[39m\n",
            "\u001b[33m543\u001b[39m\n",
            "\u001b[33m544\u001b[39m\n",
            "\u001b[33m545\u001b[39m\n",
            "\u001b[33m546\u001b[39m\n",
            "\u001b[33m547\u001b[39m\n",
            "\u001b[33m548\u001b[39m\n",
            "\u001b[33m549\u001b[39m\n",
            "\u001b[33m550\u001b[39m\n",
            "\u001b[33m551\u001b[39m\n",
            "\u001b[33m552\u001b[39m\n",
            "\u001b[33m553\u001b[39m\n",
            "\u001b[33m554\u001b[39m\n",
            "\u001b[33m555\u001b[39m\n",
            "\u001b[33m556\u001b[39m\n",
            "\u001b[33m557\u001b[39m\n",
            "\u001b[33m558\u001b[39m\n",
            "\u001b[33m559\u001b[39m\n",
            "\u001b[33m560\u001b[39m\n",
            "\u001b[33m561\u001b[39m\n",
            "\u001b[33m562\u001b[39m\n",
            "\u001b[33m563\u001b[39m\n",
            "\u001b[33m564\u001b[39m\n",
            "\u001b[33m565\u001b[39m\n",
            "\u001b[33m566\u001b[39m\n",
            "\u001b[33m567\u001b[39m\n",
            "\u001b[33m568\u001b[39m\n",
            "\u001b[33m569\u001b[39m\n",
            "\u001b[33m570\u001b[39m\n",
            "\u001b[33m571\u001b[39m\n",
            "\u001b[33m572\u001b[39m\n",
            "\u001b[33m573\u001b[39m\n",
            "\u001b[33m574\u001b[39m\n",
            "\u001b[33m575\u001b[39m\n",
            "\u001b[33m576\u001b[39m\n",
            "\u001b[33m577\u001b[39m\n",
            "\u001b[33m578\u001b[39m\n",
            "\u001b[33m579\u001b[39m\n",
            "\u001b[33m580\u001b[39m\n",
            "\u001b[33m581\u001b[39m\n",
            "\u001b[33m582\u001b[39m\n",
            "\u001b[33m583\u001b[39m\n",
            "\u001b[33m584\u001b[39m\n",
            "\u001b[33m585\u001b[39m\n",
            "\u001b[33m586\u001b[39m\n",
            "\u001b[33m587\u001b[39m\n",
            "\u001b[33m588\u001b[39m\n",
            "\u001b[33m589\u001b[39m\n",
            "\u001b[33m590\u001b[39m\n",
            "\u001b[33m591\u001b[39m\n",
            "\u001b[33m592\u001b[39m\n",
            "\u001b[33m593\u001b[39m\n",
            "\u001b[33m594\u001b[39m\n",
            "\u001b[33m595\u001b[39m\n",
            "\u001b[33m596\u001b[39m\n",
            "\u001b[33m597\u001b[39m\n",
            "\u001b[33m598\u001b[39m\n",
            "\u001b[33m599\u001b[39m\n",
            "\u001b[33m600\u001b[39m\n",
            "\u001b[33m601\u001b[39m\n",
            "\u001b[33m602\u001b[39m\n",
            "\u001b[33m603\u001b[39m\n",
            "\u001b[33m604\u001b[39m\n",
            "\u001b[33m605\u001b[39m\n",
            "\u001b[33m606\u001b[39m\n",
            "\u001b[33m607\u001b[39m\n",
            "\u001b[33m608\u001b[39m\n",
            "\u001b[33m609\u001b[39m\n",
            "\u001b[33m610\u001b[39m\n",
            "\u001b[33m611\u001b[39m\n",
            "\u001b[33m612\u001b[39m\n",
            "\u001b[33m613\u001b[39m\n",
            "\u001b[33m614\u001b[39m\n",
            "\u001b[33m615\u001b[39m\n",
            "\u001b[33m616\u001b[39m\n",
            "\u001b[33m617\u001b[39m\n",
            "\u001b[33m618\u001b[39m\n",
            "\u001b[33m619\u001b[39m\n",
            "\u001b[33m620\u001b[39m\n",
            "\u001b[33m621\u001b[39m\n",
            "\u001b[33m622\u001b[39m\n",
            "\u001b[33m623\u001b[39m\n",
            "\u001b[33m624\u001b[39m\n",
            "\u001b[33m625\u001b[39m\n",
            "\u001b[33m626\u001b[39m\n",
            "\u001b[33m627\u001b[39m\n",
            "\u001b[33m628\u001b[39m\n",
            "\u001b[33m629\u001b[39m\n",
            "\u001b[33m630\u001b[39m\n",
            "\u001b[33m631\u001b[39m\n",
            "\u001b[33m632\u001b[39m\n",
            "\u001b[33m633\u001b[39m\n",
            "\u001b[33m634\u001b[39m\n",
            "\u001b[33m635\u001b[39m\n",
            "\u001b[33m636\u001b[39m\n",
            "\u001b[33m637\u001b[39m\n",
            "\u001b[33m638\u001b[39m\n",
            "\u001b[33m639\u001b[39m\n",
            "\u001b[33m640\u001b[39m\n",
            "\u001b[33m641\u001b[39m\n",
            "\u001b[33m642\u001b[39m\n",
            "\u001b[33m643\u001b[39m\n",
            "\u001b[33m644\u001b[39m\n",
            "\u001b[33m645\u001b[39m\n",
            "\u001b[33m646\u001b[39m\n",
            "\u001b[33m647\u001b[39m\n",
            "\u001b[33m648\u001b[39m\n",
            "\u001b[33m649\u001b[39m\n",
            "\u001b[33m650\u001b[39m\n",
            "\u001b[33m651\u001b[39m\n",
            "\u001b[33m652\u001b[39m\n",
            "\u001b[33m653\u001b[39m\n",
            "\u001b[33m654\u001b[39m\n",
            "\u001b[33m655\u001b[39m\n",
            "\u001b[33m656\u001b[39m\n",
            "\u001b[33m657\u001b[39m\n",
            "\u001b[33m658\u001b[39m\n",
            "\u001b[33m659\u001b[39m\n",
            "\u001b[33m660\u001b[39m\n",
            "\u001b[33m661\u001b[39m\n",
            "\u001b[33m662\u001b[39m\n",
            "\u001b[33m663\u001b[39m\n",
            "\u001b[33m664\u001b[39m\n",
            "\u001b[33m665\u001b[39m\n",
            "\u001b[33m666\u001b[39m\n",
            "\u001b[33m667\u001b[39m\n",
            "\u001b[33m668\u001b[39m\n",
            "\u001b[33m669\u001b[39m\n",
            "\u001b[33m670\u001b[39m\n",
            "\u001b[33m671\u001b[39m\n",
            "\u001b[33m672\u001b[39m\n",
            "\u001b[33m673\u001b[39m\n",
            "\u001b[33m674\u001b[39m\n",
            "\u001b[33m675\u001b[39m\n",
            "\u001b[33m676\u001b[39m\n",
            "\u001b[33m677\u001b[39m\n",
            "\u001b[33m678\u001b[39m\n",
            "\u001b[33m679\u001b[39m\n",
            "\u001b[33m680\u001b[39m\n",
            "\u001b[33m681\u001b[39m\n",
            "\u001b[33m682\u001b[39m\n",
            "\u001b[33m683\u001b[39m\n",
            "\u001b[33m684\u001b[39m\n",
            "\u001b[33m685\u001b[39m\n",
            "\u001b[33m686\u001b[39m\n",
            "\u001b[33m687\u001b[39m\n",
            "\u001b[33m688\u001b[39m\n",
            "\u001b[33m689\u001b[39m\n",
            "\u001b[33m690\u001b[39m\n",
            "\u001b[33m691\u001b[39m\n",
            "\u001b[33m692\u001b[39m\n",
            "\u001b[33m693\u001b[39m\n",
            "\u001b[33m694\u001b[39m\n",
            "\u001b[33m695\u001b[39m\n",
            "\u001b[33m696\u001b[39m\n",
            "\u001b[33m697\u001b[39m\n",
            "\u001b[33m698\u001b[39m\n",
            "\u001b[33m699\u001b[39m\n",
            "\u001b[33m700\u001b[39m\n",
            "\u001b[33m701\u001b[39m\n",
            "\u001b[33m702\u001b[39m\n",
            "\u001b[33m703\u001b[39m\n",
            "\u001b[33m704\u001b[39m\n",
            "\u001b[33m705\u001b[39m\n",
            "\u001b[33m706\u001b[39m\n",
            "\u001b[33m707\u001b[39m\n",
            "\u001b[33m708\u001b[39m\n",
            "\u001b[33m709\u001b[39m\n",
            "\u001b[33m710\u001b[39m\n",
            "\u001b[33m711\u001b[39m\n",
            "\u001b[33m712\u001b[39m\n",
            "\u001b[33m713\u001b[39m\n",
            "\u001b[33m714\u001b[39m\n",
            "\u001b[33m715\u001b[39m\n",
            "\u001b[33m716\u001b[39m\n",
            "\u001b[33m717\u001b[39m\n",
            "\u001b[33m718\u001b[39m\n",
            "\u001b[33m719\u001b[39m\n",
            "\u001b[33m720\u001b[39m\n",
            "\u001b[33m721\u001b[39m\n",
            "\u001b[33m722\u001b[39m\n",
            "\u001b[33m723\u001b[39m\n",
            "\u001b[33m724\u001b[39m\n",
            "\u001b[33m725\u001b[39m\n",
            "\u001b[33m726\u001b[39m\n",
            "\u001b[33m727\u001b[39m\n",
            "\u001b[33m728\u001b[39m\n",
            "\u001b[33m729\u001b[39m\n",
            "\u001b[33m730\u001b[39m\n",
            "\u001b[33m731\u001b[39m\n",
            "\u001b[33m732\u001b[39m\n",
            "\u001b[33m733\u001b[39m\n",
            "\u001b[33m734\u001b[39m\n",
            "\u001b[33m735\u001b[39m\n",
            "\u001b[33m736\u001b[39m\n",
            "\u001b[33m737\u001b[39m\n",
            "\u001b[33m738\u001b[39m\n",
            "\u001b[33m739\u001b[39m\n",
            "\u001b[33m740\u001b[39m\n",
            "\u001b[33m741\u001b[39m\n",
            "\u001b[33m742\u001b[39m\n",
            "\u001b[33m743\u001b[39m\n",
            "\u001b[33m744\u001b[39m\n",
            "\u001b[33m745\u001b[39m\n",
            "\u001b[33m746\u001b[39m\n",
            "\u001b[33m747\u001b[39m\n",
            "\u001b[33m748\u001b[39m\n",
            "\u001b[33m749\u001b[39m\n",
            "\u001b[33m750\u001b[39m\n",
            "\u001b[33m751\u001b[39m\n",
            "\u001b[33m752\u001b[39m\n",
            "\u001b[33m753\u001b[39m\n",
            "\u001b[33m754\u001b[39m\n",
            "\u001b[33m755\u001b[39m\n",
            "\u001b[33m756\u001b[39m\n",
            "\u001b[33m757\u001b[39m\n",
            "\u001b[33m758\u001b[39m\n",
            "\u001b[33m759\u001b[39m\n",
            "\u001b[33m760\u001b[39m\n",
            "\u001b[33m761\u001b[39m\n",
            "\u001b[33m762\u001b[39m\n",
            "\u001b[33m763\u001b[39m\n",
            "\u001b[33m764\u001b[39m\n",
            "\u001b[33m765\u001b[39m\n",
            "\u001b[33m766\u001b[39m\n",
            "\u001b[33m767\u001b[39m\n",
            "\u001b[33m768\u001b[39m\n",
            "\u001b[33m769\u001b[39m\n",
            "\u001b[33m770\u001b[39m\n",
            "\u001b[33m771\u001b[39m\n",
            "\u001b[33m772\u001b[39m\n",
            "\u001b[33m773\u001b[39m\n",
            "\u001b[33m774\u001b[39m\n",
            "\u001b[33m775\u001b[39m\n",
            "\u001b[33m776\u001b[39m\n",
            "\u001b[33m777\u001b[39m\n",
            "\u001b[33m778\u001b[39m\n",
            "\u001b[33m779\u001b[39m\n",
            "\u001b[33m780\u001b[39m\n",
            "\u001b[33m781\u001b[39m\n",
            "\u001b[33m782\u001b[39m\n",
            "\u001b[33m783\u001b[39m\n",
            "\u001b[33m784\u001b[39m\n",
            "\u001b[33m785\u001b[39m\n",
            "\u001b[33m786\u001b[39m\n",
            "\u001b[33m787\u001b[39m\n",
            "\u001b[33m788\u001b[39m\n",
            "\u001b[33m789\u001b[39m\n",
            "\u001b[33m790\u001b[39m\n",
            "\u001b[33m791\u001b[39m\n",
            "\u001b[33m792\u001b[39m\n",
            "\u001b[33m793\u001b[39m\n",
            "\u001b[33m794\u001b[39m\n",
            "\u001b[33m795\u001b[39m\n",
            "\u001b[33m796\u001b[39m\n",
            "\u001b[33m797\u001b[39m\n",
            "\u001b[33m798\u001b[39m\n",
            "\u001b[33m799\u001b[39m\n",
            "\u001b[33m800\u001b[39m\n",
            "\u001b[33m801\u001b[39m\n",
            "\u001b[33m802\u001b[39m\n",
            "\u001b[33m803\u001b[39m\n",
            "\u001b[33m804\u001b[39m\n",
            "\u001b[33m805\u001b[39m\n",
            "\u001b[33m806\u001b[39m\n",
            "\u001b[33m807\u001b[39m\n",
            "\u001b[33m808\u001b[39m\n",
            "\u001b[33m809\u001b[39m\n",
            "\u001b[33m810\u001b[39m\n",
            "\u001b[33m811\u001b[39m\n",
            "\u001b[33m812\u001b[39m\n",
            "\u001b[33m813\u001b[39m\n",
            "\u001b[33m814\u001b[39m\n",
            "\u001b[33m815\u001b[39m\n",
            "\u001b[33m816\u001b[39m\n",
            "\u001b[33m817\u001b[39m\n",
            "\u001b[33m818\u001b[39m\n",
            "\u001b[33m819\u001b[39m\n",
            "\u001b[33m820\u001b[39m\n",
            "\u001b[33m821\u001b[39m\n",
            "\u001b[33m822\u001b[39m\n",
            "\u001b[33m823\u001b[39m\n",
            "\u001b[33m824\u001b[39m\n",
            "\u001b[33m825\u001b[39m\n",
            "\u001b[33m826\u001b[39m\n",
            "\u001b[33m827\u001b[39m\n",
            "\u001b[33m828\u001b[39m\n",
            "\u001b[33m829\u001b[39m\n",
            "\u001b[33m830\u001b[39m\n",
            "\u001b[33m831\u001b[39m\n",
            "\u001b[33m832\u001b[39m\n",
            "\u001b[33m833\u001b[39m\n",
            "\u001b[33m834\u001b[39m\n",
            "\u001b[33m835\u001b[39m\n",
            "\u001b[33m836\u001b[39m\n",
            "\u001b[33m837\u001b[39m\n",
            "\u001b[33m838\u001b[39m\n",
            "\u001b[33m839\u001b[39m\n",
            "\u001b[33m840\u001b[39m\n",
            "\u001b[33m841\u001b[39m\n",
            "\u001b[33m842\u001b[39m\n",
            "\u001b[33m843\u001b[39m\n",
            "\u001b[33m844\u001b[39m\n",
            "\u001b[33m845\u001b[39m\n",
            "\u001b[33m846\u001b[39m\n",
            "\u001b[33m847\u001b[39m\n",
            "\u001b[33m848\u001b[39m\n",
            "\u001b[33m849\u001b[39m\n",
            "\u001b[33m850\u001b[39m\n",
            "\u001b[33m851\u001b[39m\n",
            "\u001b[33m852\u001b[39m\n",
            "\u001b[33m853\u001b[39m\n",
            "\u001b[33m854\u001b[39m\n",
            "\u001b[33m855\u001b[39m\n",
            "\u001b[33m856\u001b[39m\n",
            "\u001b[33m857\u001b[39m\n",
            "\u001b[33m858\u001b[39m\n",
            "\u001b[33m859\u001b[39m\n",
            "\u001b[33m860\u001b[39m\n",
            "\u001b[33m861\u001b[39m\n",
            "\u001b[33m862\u001b[39m\n",
            "\u001b[33m863\u001b[39m\n",
            "\u001b[33m864\u001b[39m\n",
            "\u001b[33m865\u001b[39m\n",
            "\u001b[33m866\u001b[39m\n",
            "\u001b[33m867\u001b[39m\n",
            "\u001b[33m868\u001b[39m\n",
            "\u001b[33m869\u001b[39m\n",
            "\u001b[33m870\u001b[39m\n",
            "\u001b[33m871\u001b[39m\n",
            "\u001b[33m872\u001b[39m\n",
            "\u001b[33m873\u001b[39m\n",
            "\u001b[33m874\u001b[39m\n",
            "\u001b[33m875\u001b[39m\n",
            "\u001b[33m876\u001b[39m\n",
            "\u001b[33m877\u001b[39m\n",
            "\u001b[33m878\u001b[39m\n",
            "\u001b[33m879\u001b[39m\n",
            "\u001b[33m880\u001b[39m\n",
            "\u001b[33m881\u001b[39m\n",
            "\u001b[33m882\u001b[39m\n",
            "\u001b[33m883\u001b[39m\n",
            "\u001b[33m884\u001b[39m\n",
            "\u001b[33m885\u001b[39m\n",
            "\u001b[33m886\u001b[39m\n",
            "\u001b[33m887\u001b[39m\n",
            "\u001b[33m888\u001b[39m\n",
            "\u001b[33m889\u001b[39m\n",
            "\u001b[33m890\u001b[39m\n",
            "\u001b[33m891\u001b[39m\n",
            "\u001b[33m892\u001b[39m\n",
            "\u001b[33m893\u001b[39m\n",
            "\u001b[33m894\u001b[39m\n",
            "\u001b[33m895\u001b[39m\n",
            "\u001b[33m896\u001b[39m\n",
            "\u001b[33m897\u001b[39m\n",
            "\u001b[33m898\u001b[39m\n",
            "\u001b[33m899\u001b[39m\n",
            "\u001b[33m900\u001b[39m\n",
            "\u001b[33m901\u001b[39m\n",
            "\u001b[33m902\u001b[39m\n",
            "\u001b[33m903\u001b[39m\n",
            "\u001b[33m904\u001b[39m\n",
            "\u001b[33m905\u001b[39m\n",
            "\u001b[33m906\u001b[39m\n",
            "\u001b[33m907\u001b[39m\n",
            "\u001b[33m908\u001b[39m\n",
            "\u001b[33m909\u001b[39m\n",
            "\u001b[33m910\u001b[39m\n",
            "\u001b[33m911\u001b[39m\n",
            "\u001b[33m912\u001b[39m\n",
            "\u001b[33m913\u001b[39m\n",
            "\u001b[33m914\u001b[39m\n",
            "\u001b[33m915\u001b[39m\n",
            "\u001b[33m916\u001b[39m\n",
            "\u001b[33m917\u001b[39m\n",
            "\u001b[33m918\u001b[39m\n",
            "\u001b[33m919\u001b[39m\n",
            "\u001b[33m920\u001b[39m\n",
            "\u001b[33m921\u001b[39m\n",
            "\u001b[33m922\u001b[39m\n",
            "\u001b[33m923\u001b[39m\n",
            "\u001b[33m924\u001b[39m\n",
            "\u001b[33m925\u001b[39m\n",
            "\u001b[33m926\u001b[39m\n",
            "\u001b[33m927\u001b[39m\n",
            "\u001b[33m928\u001b[39m\n",
            "\u001b[33m929\u001b[39m\n",
            "\u001b[33m930\u001b[39m\n",
            "\u001b[33m931\u001b[39m\n",
            "\u001b[33m932\u001b[39m\n",
            "\u001b[33m933\u001b[39m\n",
            "\u001b[33m934\u001b[39m\n",
            "\u001b[33m935\u001b[39m\n",
            "\u001b[33m936\u001b[39m\n",
            "\u001b[33m937\u001b[39m\n",
            "\u001b[33m938\u001b[39m\n",
            "\u001b[33m939\u001b[39m\n",
            "\u001b[33m940\u001b[39m\n",
            "\u001b[33m941\u001b[39m\n",
            "\u001b[33m942\u001b[39m\n",
            "\u001b[33m943\u001b[39m\n",
            "\u001b[33m944\u001b[39m\n",
            "\u001b[33m945\u001b[39m\n",
            "\u001b[33m946\u001b[39m\n",
            "\u001b[33m947\u001b[39m\n",
            "\u001b[33m948\u001b[39m\n",
            "\u001b[33m949\u001b[39m\n",
            "\u001b[33m950\u001b[39m\n",
            "\u001b[33m951\u001b[39m\n",
            "\u001b[33m952\u001b[39m\n",
            "\u001b[33m953\u001b[39m\n",
            "\u001b[33m954\u001b[39m\n",
            "\u001b[33m955\u001b[39m\n",
            "\u001b[33m956\u001b[39m\n",
            "\u001b[33m957\u001b[39m\n",
            "\u001b[33m958\u001b[39m\n",
            "\u001b[33m959\u001b[39m\n",
            "\u001b[33m960\u001b[39m\n",
            "\u001b[33m961\u001b[39m\n",
            "\u001b[33m962\u001b[39m\n",
            "\u001b[33m963\u001b[39m\n",
            "\u001b[33m964\u001b[39m\n",
            "\u001b[33m965\u001b[39m\n",
            "\u001b[33m966\u001b[39m\n",
            "\u001b[33m967\u001b[39m\n",
            "\u001b[33m968\u001b[39m\n",
            "\u001b[33m969\u001b[39m\n",
            "\u001b[33m970\u001b[39m\n",
            "\u001b[33m971\u001b[39m\n",
            "\u001b[33m972\u001b[39m\n",
            "\u001b[33m973\u001b[39m\n",
            "\u001b[33m974\u001b[39m\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Running the method using LeRes\n",
        "!python run.py --Final --data_dir {INPUT_FILES} --output_dir  {OUTPUT_FILES}/ --depthNet 2"
      ],
      "metadata": {
        "id": "gDgQ4X_ubiNP",
        "outputId": "b1fcee74-8cb6-40ff-a2c5-a193b49e6dad",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "device: cuda\n",
            "Namespace(data_dir='/content/input1', output_dir='/content/outputs_leres/', savepatchs=0, savewholeest=0, output_resolution=1, net_receptive_field_size=None, pix2pixsize=1024, depthNet=2, colorize_results=False, R0=False, R20=False, Final=True, max_res=inf)\n",
            "----------------- Options ---------------\n",
            "                    Final: True                          \t[default: False]\n",
            "                       R0: False                         \n",
            "                      R20: False                         \n",
            "             aspect_ratio: 1.0                           \n",
            "               batch_size: 1                             \n",
            "          checkpoints_dir: ./pix2pix/checkpoints         \n",
            "         colorize_results: False                         \n",
            "                crop_size: 672                           \n",
            "                 data_dir: /content/input1               \t[default: None]\n",
            "                 dataroot: None                          \n",
            "             dataset_mode: depthmerge                    \n",
            "                 depthNet: 2                             \t[default: None]\n",
            "                direction: AtoB                          \n",
            "          display_winsize: 256                           \n",
            "                    epoch: latest                        \n",
            "                     eval: False                         \n",
            "            generatevideo: None                          \n",
            "                  gpu_ids: 0                             \n",
            "                init_gain: 0.02                          \n",
            "                init_type: normal                        \n",
            "                 input_nc: 2                             \n",
            "                  isTrain: False                         \t[default: None]\n",
            "                load_iter: 0                             \t[default: 0]\n",
            "                load_size: 672                           \n",
            "         max_dataset_size: 10000                         \n",
            "                  max_res: inf                           \n",
            "                    model: pix2pix4depth                 \n",
            "               n_layers_D: 3                             \n",
            "                     name: void                          \n",
            "                      ndf: 64                            \n",
            "                     netD: basic                         \n",
            "                     netG: unet_1024                     \n",
            " net_receptive_field_size: None                          \n",
            "                      ngf: 64                            \n",
            "               no_dropout: False                         \n",
            "                  no_flip: False                         \n",
            "                     norm: none                          \n",
            "                 num_test: 50                            \n",
            "              num_threads: 4                             \n",
            "               output_dir: /content/outputs_leres/       \t[default: None]\n",
            "                output_nc: 1                             \n",
            "        output_resolution: None                          \n",
            "                    phase: test                          \n",
            "              pix2pixsize: None                          \n",
            "               preprocess: resize_and_crop               \n",
            "                savecrops: None                          \n",
            "             savewholeest: None                          \n",
            "           serial_batches: False                         \n",
            "                   suffix:                               \n",
            "                  verbose: False                         \n",
            "----------------- End -------------------\n",
            "Traceback (most recent call last):\n",
            "  File \"/content/BoostingMonocularDepth/run.py\", line 575, in <module>\n",
            "    run(dataset_, option_)\n",
            "  File \"/content/BoostingMonocularDepth/run.py\", line 48, in run\n",
            "    opt = TestOptions().parse()\n",
            "  File \"/content/BoostingMonocularDepth/pix2pix/options/base_options.py\", line 152, in parse\n",
            "    torch.cuda.set_device(opt.gpu_ids[0])\n",
            "  File \"/usr/local/lib/python3.10/dist-packages/torch/cuda/__init__.py\", line 404, in set_device\n",
            "    torch._C._cuda_setDevice(device)\n",
            "  File \"/usr/local/lib/python3.10/dist-packages/torch/cuda/__init__.py\", line 298, in _lazy_init\n",
            "    torch._C._cuda_init()\n",
            "RuntimeError: Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "**Create Video Output**"
      ],
      "metadata": {
        "id": "ZIBs_0aDbq46"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "#create video 1\n",
        "!ffmpeg \\\n",
        "-framerate 24 \\\n",
        "-pattern_type glob \\\n",
        "-i {INPUT_FILES}'/*.png' \\\n",
        "input0.mp4\n",
        "#create vodeo 2\n",
        "!ffmpeg \\\n",
        "-framerate 24 \\\n",
        "-pattern_type glob \\\n",
        "-i {OUTPUT_FILES}'/*.png' \\\n",
        "input1.mp4\n",
        "#merge video 1 and video 2\n",
        "!ffmpeg \\\n",
        "-i input0.mp4 -i input1.mp4 \\\n",
        "-filter_complex \\\n",
        "\"[0:v][1:v]vstack=inputs=2\" \\\n",
        "finalOutput.mp4\n",
        "#add sound of main video to final output video\n",
        "!ffmpeg \\\n",
        "-i 'finalOutput.mp4' -i 'm.mp4' \\\n",
        "-c:v copy -c:a copy -map 0:v:0 -map 1:a:0 \\\n",
        "OUTPUT_FILE.mp4"
      ],
      "metadata": {
        "id": "IHv3_wz1bs4_"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "jzkIeHiNc9Zv"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "metadata": {
        "id": "pd0LfKcGizVs"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "%cd /content/drive/MyDrive/OUT/data/machine_learning_test_dataset/drive"
      ],
      "metadata": {
        "id": "rPX-Npz7qQYs"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "!pwd"
      ],
      "metadata": {
        "id": "EF4kDZevqkmi"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "**FFMPEG video to image**"
      ],
      "metadata": {
        "id": "BpmpJgkHa0fy"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "##**MLP MODEL WITH POSITIONAL ENCODER**"
      ],
      "metadata": {
        "id": "UFAs21IJafFS"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "icmWdrOM9Rx1"
      },
      "outputs": [],
      "source": [
        "############<< LOADING DATASET >>#############\n",
        "\n",
        "!cp \"/content/drive/MyDrive/OUT/data/machine_learning_test_dataset/archive.zip\" \"/content/\"\n",
        "!unzip /content/archive.zip #unzipping the flower images from archive..\n",
        "!mkdir -p /content/flowers/all\n",
        "!cp /content/flowers/daisy/* /content/flowers/all\n",
        "!cp /content/flowers/dandelion/* /content/flowers/all\n",
        "!cp /content/flowers/rose/* /content/flowers/all\n",
        "!cp /content/flowers/sunflower/* /content/flowers/all\n",
        "!cp /content/flowers/tulip/* /content/flowers/all\n",
        "\n",
        "##############$$$$$$$$$$$$$$<<< IMPORTING LIBRARIES >>>$$$$$$$$$$$$$$$$################\n",
        "#✅\n",
        "import jax\n",
        "import jax.numpy as jnp\n",
        "\n",
        "!python -m pip install -q -U flax\n",
        "import flax\n",
        "import optax\n",
        "from typing import Any\n",
        "\n",
        "from jax import lax\n",
        "import flax.linen as nn\n",
        "from flax.training import train_state\n",
        "##########################################<< FFT PSEUDO SIN COS POSITIONAL ENCODER >>#########################################\n",
        "\n",
        "positional_encoding_dims = 6  # Number of positional encodings applied\n",
        "\n",
        "def positional_encoding(inputs):\n",
        "    print(\"positional_encoding start\")\n",
        "    batch_size, _ = inputs.shape;print(inputs.shape)\n",
        "    inputs_freq = jax.vmap(lambda x: inputs * 2.0 ** x)(jnp.arange(positional_encoding_dims));print(inputs_freq.shape)\n",
        "    x = jnp.stack([jnp.sin(inputs_freq), jnp.cos(inputs_freq)]);print(x.shape)\n",
        "    x = x.swapaxes(0, 2);print(x.shape)\n",
        "    x = x.reshape([batch_size, -1]);print(x.shape)\n",
        "    x = jnp.concatenate([inputs, x], axis=-1);print(x.shape)\n",
        "    print(\"positional_encoding end\")\n",
        "    return x\n",
        "##########################################<< MLP MODEL >>#########################################\n",
        "\n",
        "apply_positional_encoding = True # Apply posittional encoding to the input or not\n",
        "num_dense_layers = 8 # Number of dense layers in MLP\n",
        "dense_layer_width = 256 # Dimentionality of dense layers' output space\n",
        "\n",
        "class MLPModel(nn.Module):\n",
        "    dtype: Any = jnp.float32\n",
        "    precision: Any = lax.Precision.DEFAULT\n",
        "    apply_positional_encoding: bool = apply_positional_encoding\n",
        "    @nn.compact\n",
        "    def __call__(self, input_points):\n",
        "        x = positional_encoding(input_points) if self.apply_positional_encoding else input_points\n",
        "        print(\"network model start\")\n",
        "        print(x.shape)\n",
        "        for i in range(num_dense_layers):\n",
        "            x = nn.Dense(\n",
        "                dense_layer_width,\n",
        "                dtype=self.dtype,\n",
        "                precision=self.precision\n",
        "                )(x)\n",
        "            x = nn.relu(x)\n",
        "            x = jnp.concatenate([x, input_points], axis=-1) if i == 4 else x\n",
        "            print(x.shape)\n",
        "\n",
        "        x = nn.Dense(1, dtype=self.dtype, precision=self.precision)(x)\n",
        "        print(x.shape)\n",
        "        print(\"network model end\")\n",
        "        return x\n",
        "##########################################<< MLP MODEL >>#########################################\n",
        "\n",
        "##########################################<< CREATE MODEL INITIAL STATE FILLING WITH WEIGHTS >>#########################################\n",
        "\n",
        "def init_train_state(model, r_key, shape, learning_rate ) -> train_state.TrainState:\n",
        "    print(shape)\n",
        "    init_variables = model.init(r_key, jnp.ones(shape))  # Initialize the Model\n",
        "    optimizer = optax.adam(learning_rate) # Create the optimizer\n",
        "    # Create a State\n",
        "    return train_state.TrainState.create(\n",
        "        apply_fn = model.apply,\n",
        "        tx=optimizer,\n",
        "        params=init_variables['params']\n",
        "    )\n",
        "\n",
        "learning_rate = 1e-4\n",
        "batch_size_no = 64\n",
        "\n",
        "model = MLPModel() # Instantiate the Model\n",
        "\n",
        "key, rng = jax.random.split(jax.random.PRNGKey(0))\n",
        "x = jnp.empty((batch_size_no, 28, 28, 1)) # Dummy Input\n",
        "_, image_height, image_width, channels = x.shape\n",
        "##########################################<< CREATING MODEL STATE>>#########################################\n",
        "\n",
        "state = init_train_state( model, rng, (image_height * image_width, channels), learning_rate )\n",
        "state\n",
        "print(\"state initiated ✅\")\n",
        "\n",
        "##########################################<< MODEL LOSS METRICS FOR GRADIENT COMPUTATION >>#########################################\n",
        "def image_difference_loss(logits, labels):\n",
        "    loss = .5 * jnp.mean((logits - labels) ** 2)\n",
        "    return loss\n",
        "def compute_metrics(*, logits, labels):\n",
        "  loss = image_difference_loss(logits, labels)\n",
        "  loss = lax.pmean(loss, axis_name=\"batch\");print(\"ok4\")\n",
        "  metrics = {\n",
        "      'loss': loss,\n",
        "      'logits': logits, #PREDICTED IMAGE\n",
        "      'labels': labels  #ACTUAL IMAGE\n",
        "  }\n",
        "  return metrics\n",
        "\n",
        "##########################################<< MODEL TRAINING STEP FOR UPDATING WEIGHTS FROM CALCULATED GRAGIENTS FROM CHANGING MODEL STATE >>#########################################\n",
        "import jax\n",
        "\n",
        "def train_step(state: train_state.TrainState, batch: jnp.asarray, rng):\n",
        "    print(batch)\n",
        "    image, label = batch\n",
        "    print(image,\"<<<image\")\n",
        "    print(label,\"<<<label\")\n",
        "    def loss_fn(params):\n",
        "        logits = state.apply_fn({'params': params}, image);print(\"done1\",logits.shape)\n",
        "        loss =  image_difference_loss(logits, label);print(\"done2\",loss.shape)\n",
        "        return loss, logits\n",
        "\n",
        "    print(\"ok1really\")\n",
        "    gradient_fn = jax.value_and_grad(loss_fn, has_aux=True);print(\"ok1\")\n",
        "    (_, logits), grads = gradient_fn(state.params);print(\"ok2\")\n",
        "    #train_loss, gradients_each = jax.value_and_grad(loss_fn)(state.params);print(\"ok3\")\n",
        "    grads = lax.pmean(grads,\"batch\");print(\"ok4\")\n",
        "    # grads = jnp.mean(grads);print(\"ok4\")\n",
        "    state = state.apply_gradients(grads=grads);print(\"ok5\")\n",
        "    # train_loss = jnp.mean(train_loss);print(\"ok6\")\n",
        "    logs = compute_metrics(logits=logits, labels=label);print(\"ok7\")\n",
        "    return state, logs\n",
        "\n",
        "parallel_train_step = jax.pmap(train_step, \"batch\")\n",
        "# parallel_train_step = jax.pmap(train_step, axis_name=\"batch\", in_axes = (0, 0, 0))\n",
        "\n",
        "import jax\n",
        "@jax.jit\n",
        "def eval_step(state, batch):\n",
        "    image, label = batch\n",
        "    logits = state.apply_fn({'params': state.params}, image)\n",
        "    return compute_metrics(logits=logits, labels=label)\n",
        "\n",
        "##########################################<< DATASET GENERATING FUNCTION HELPER IN GRB TO GRAY CONVERSION >>#########################################\n",
        "newsize = (150, 150) # /.... 233 * 454\n",
        "from PIL import Image\n",
        "import jax.numpy as jnp\n",
        "\n",
        "def imageGRAY(argv):\n",
        "    im = Image.open(argv).convert('L')\n",
        "    tvt, tvu = jnp.asarray(im.resize(newsize)),jnp.asarray(im.resize(newsize)).reshape(-1,1)\n",
        "    return tvt, tvu\n",
        "def imageRGB(argv):\n",
        "    im = Image.open(argv)\n",
        "    tvt, tvu = jnp.asarray(im.resize(newsize)),jnp.asarray(im.resize(newsize)).reshape(-1,3)\n",
        "    return tvt, tvu\n",
        "\n",
        "\n",
        "##########################################<< ANALYSING DATASETS FOR BATCH DISTRIBUTION >>#########################################\n",
        "# testing batching of the dataset frfom the total dataset, by assuming batch size as 8 then running 50 epochs over the batch , then moving processing onto next batch>>>\n",
        "batch_size = jax.device_count()    #///// batchsize depending on number of devices available for processing\n",
        "import os\n",
        "image_dir = r'/content/flowers/all/'\n",
        "#############################################################################\n",
        "prefix = \"sur_refl_\"\n",
        "bandend = [\"c\",\".jpg\", \"b02\", \"b03\", \"b04\", \"b05\", \"b06\", \"b07\", \"day_of_year\", \"qc_500m\", \"raz\", \"state_500m\", \"szen\", \"vzen\"]\n",
        "DayOY = \"_doy\\[0-9]+_aid0001\"\n",
        "fileExt = r'.jpg'\n",
        "expression_b2 = bandend[1]\n",
        "total_images =  [f for f in os.listdir(image_dir) if f.__contains__(expression_b2)]\n",
        "total_images.sort()\n",
        "total_images_path = [os.path.join(image_dir, i) for i in total_images if i != 'outputs']\n",
        "no_of_batches = int(len(total_images_path)/batch_size)\n",
        "\n",
        "\n",
        "\n",
        "##########################################<< RUNNING TRAINNING >>#########################################\n",
        "#@title # **👠HIGH HEELS RUN >>>>>>>>>>>** { vertical-output: true }\n",
        "newsize = (260, 260) # /.... 233 * 454\n",
        "\n",
        "import jax\n",
        "from jax import random\n",
        "def batchedimages(image_locations):\n",
        "  ddyss = jnp.asarray((imageRGB(total_images_path[image_locations[0]])[1],\n",
        "                      imageRGB(total_images_path[image_locations[1]])[1],\n",
        "                      imageRGB(total_images_path[image_locations[2]])[1],\n",
        "                      imageRGB(total_images_path[image_locations[3]])[1],\n",
        "                      imageRGB(total_images_path[image_locations[4]])[1],\n",
        "                      imageRGB(total_images_path[image_locations[5]])[1],\n",
        "                      imageRGB(total_images_path[image_locations[6]])[1],\n",
        "                      imageRGB(total_images_path[image_locations[7]])[1]))\n",
        "  ddxss = jnp.asarray((imageGRAY(total_images_path[image_locations[0]])[1],\n",
        "                      imageGRAY(total_images_path[image_locations[1]])[1],\n",
        "                      imageGRAY(total_images_path[image_locations[2]])[1],\n",
        "                      imageGRAY(total_images_path[image_locations[3]])[1],\n",
        "                      imageGRAY(total_images_path[image_locations[4]])[1],\n",
        "                      imageGRAY(total_images_path[image_locations[5]])[1],\n",
        "                      imageGRAY(total_images_path[image_locations[6]])[1],\n",
        "                      imageGRAY(total_images_path[image_locations[7]])[1]))\n",
        "  #print(ddyss.shape,\"<<<< ddyss.shape???\",ddxss.shape,\"<<<< ddxss.shape???\") #to check shape\n",
        "  batch_ccc = ddyss, ddxss\n",
        "  return batch_ccc\n",
        "\n",
        "def data_stream():\n",
        "  key = random.PRNGKey(0)\n",
        "  perm = random.permutation(key, len(total_images_path))\n",
        "  for i in range(no_of_batches):\n",
        "    batch_idx = perm[i * batch_size : (i + 1) * batch_size]; #print(batch_idx)\n",
        "    yield batchedimages(batch_idx)\n",
        "\n",
        "batches = data_stream()  ### this stream will utilize the array of paths of images to a folder, then \"generate\" batches into the variable\n",
        "\n",
        "next(batches)[0].shape ### this command starts initial 8 image  stream, if callled inside a iteration loop then it will get next images for calculations>>>\n",
        "vv, shapea, channels = next(batches)[0].shape # seitting values >>> [0] 8 784 3  [1] 8 784 1 ; RGB & GRAYSCALE versions 8 images each converted to 1-D array\n",
        "print(vv, shapea, channels)\n",
        "######################<<< summary writer for tensor board\n",
        "# from torch.utils.tensorboard import SummaryWriter\n",
        "# logdir = \"runs\"\n",
        "# writer = SummaryWriter(logdir)\n",
        "######################\n",
        "######################\n",
        "rng = jax.random.PRNGKey(0)\n",
        "# dropout_rngs = jax.random.split(rng, jax.local_device_count())\n",
        "######################\n",
        "#################################<<< checking if checkpoint already available\n",
        "import os # importing os module\n",
        "import re # to find file using regular expression\n",
        "checkpoint_available = 0\n",
        "pattern = re.compile(\"checkpoint_\\d+\")   # to search for \"checkpoint_*munerical value*\" numerical value of any length is denoted by regular expression \"\\d+\"\n",
        "dir = \"/content/ckpts/\"\n",
        "isFile = os.path.isdir(dir)\n",
        "if isFile:\n",
        "  for filepath in os.listdir(dir):\n",
        "      if pattern.match(filepath):\n",
        "          checkpoint_available = 1\n",
        "#################################\n",
        "##########################################<<< loading checkpoint by checking the Flag available\n",
        "from flax.training import checkpoints\n",
        "if checkpoint_available:\n",
        "  CKPT_DIR = 'ckpts'\n",
        "  restored_state = checkpoints.restore_checkpoint(ckpt_dir=CKPT_DIR, target=state)\n",
        "  #state = flax.jax_utils.replicate(restored_state)\n",
        "  print(\"true <<< File loaded for and replicated to all devices\")\n",
        "##########################################\n",
        "######################<<<< initiating train state\n",
        "count = 0\n",
        "if count == 0 :\n",
        "  state = init_train_state( model, rng, (shapea, channels), learning_rate )\n",
        "  count = 1\n",
        "state = flax.jax_utils.replicate(state)  # FLAX will replicate the state to every device so that updating can be made easy\n",
        "dropout_rngs = jax.random.split(rng, jax.local_device_count())\n",
        "######################\n",
        "from flax.training import checkpoints\n",
        "import time\n",
        "total_epochs = 50\n",
        "for epochs in range(total_epochs):   # EPOCHS for training & updating the initiated state, metrics may show the loss in each epochs or iteration\n",
        "  start_time = time.time()\n",
        "  batches = data_stream()  ### this stream will utilize the array of paths of images to a folder, then \"generate\" batches into the variable\n",
        "  if checkpoint_available:\n",
        "    CKPT_DIR = 'ckpts'\n",
        "    restored_state = checkpoints.restore_checkpoint(ckpt_dir=CKPT_DIR, target=state)\n",
        "    state = restored_state\n",
        "    checkpoint_available = 0 # << Flag updated >>> to stop loading the same checkpoint in the next iteration then remove the checkpoint directory\n",
        "    !rm -r /content/ckpts\n",
        "  for bbb in range(no_of_batches-5):\n",
        "    print(bbb,\"of total number of batches\",no_of_batches)\n",
        "    state, metrics = parallel_train_step(state, next(batches), dropout_rngs)\n",
        "    print(\"<<✅✅✅epoc : \",epochs,\" complete✅✅✅>>\\n\",metrics['loss'][0]) #printing loss 1 out of 8 processed in 8 devices\n",
        "\n",
        "    #############################################<<< output visualization 👍🏻\n",
        "    from google.colab.patches import cv2_imshow\n",
        "    import numpy as np        # source : https://www.folkstalk.com/tech/how-to-convert-numpy-array-to-cv2-image-with-code-examples/\n",
        "\n",
        "    print(\"logits shape ⚡⚡⚡⚡\", metrics['logits'][0].shape)\n",
        "    L1 = metrics['logits'][0]\n",
        "    # predicted_image = np.array(L1,  dtype=np.uint8).reshape(newsize) # This would be your image array\n",
        "    # cv2img = cv2_imshow(predicted_image) # This work the same as passing an image\n",
        "    for i in range(0,metrics['labels'].shape[0]):\n",
        "      print(i)\n",
        "      predicted_image = np.array(metrics['logits'][i],  dtype=np.uint8).reshape(newsize) # This would be your image array\n",
        "      cv2_imshow(predicted_image)\n",
        "\n",
        "      print(\"labels shape ⚡⚡⚡⚡\", metrics['labels'][0].shape)\n",
        "      L2 = metrics['labels'][0]\n",
        "      actual_image = np.array(metrics['labels'][i],  dtype=np.uint8).reshape(newsize) # This would be your image array newsize = image_width , image_height\n",
        "      cv2img = cv2_imshow(actual_image) # This work the same as passing an image\n",
        "    #############################################<<< output visualization 👍🏻\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "    writer.add_scalar('Loss', int(metrics['loss'][0]), epochs)\n",
        "  epoch_time = time.time() - start_time\n",
        "  print(f\"Epoch {epochs} in {epoch_time:0.2f} sec\")\n",
        "  ##################################################<<< model saving mechanism for flax model state as checkpoints for each epochs,\"checkpoint\" is a terminology that means all the model weights and biases during the calculation till the completion of 1 epoch were being updated, then this final set of weights and biases including their placement inside the model will be saves as a (schema+weight values) saved as checkpoint in the mentioned <<CKPT_DIR = 'ckpts'>> mentioned folder.\n",
        "  CKPT_DIR = 'ckpts'\n",
        "  checkpoints.save_checkpoint(ckpt_dir=CKPT_DIR, target=state, step= epochs)     # naming of the checkpoint is \"checkpoint_*\"  where \"*\" => value of the steps variable, i.e. 'epochs'\n",
        "  restored_state = checkpoints.restore_checkpoint(ckpt_dir=CKPT_DIR, target=state) # using to get the checkpoint loaded , it can be latest one , or if already available as checkpoint in the \"CKPT_DIR\" directory then take the file from directory then save in >> restored_checkpoints\n",
        "  ##################################################\n",
        "  # images = total_images_path[batch_idx]\n",
        "writer.flush()\n",
        "\n"
      ]
    }
  ]
}