clevr_data_generation.ipynb

clevr_data_generation.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "clevr_data_generation.ipynb",
      "private_outputs": true,
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyOiaq0p/OYpNIjNqEMZTVCj",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/gist/nanlliu/ccaa9beacfbb627b2da15790c0275fa6/clevr_data_generation.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "y0RtoucqCRZc"
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/gdrive')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iumxS_FEDL3q"
      },
      "source": [
        "%cd /content/gdrive/MyDrive/clevr-dataset-gen"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ErCuPfVSEpgp"
      },
      "source": [
        "%cd /content/gdrive/MyDrive/clevr-dataset-gen/image_generation"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NT-LW2iBQ2sJ"
      },
      "source": [
        "!apt install blender\n",
        "!apt install libboost-all-dev\n",
        "!apt install libgl1-mesa-dev"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ojIwubdnROwZ"
      },
      "source": [
        "blender_version = 'blender2.79' #@param [\"blender2.79\", \"blender2.80\", \"blender2.81\", \"blender2.82\", \"blender2.83\", \"blender2.90.1\", \"blender2.91.2\"] {allow-input: false}\n",
        "gpu_enabled = True #@param {type:\"boolean\"}\n",
        "cpu_enabled = False #@param {type:\"boolean\"}"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BxAsAPz9SBUa"
      },
      "source": [
        "if blender_version == \"blender2.79\":\n",
        "    download_path=\"https://download.blender.org/release/Blender2.79/blender-2.79-linux-glibc219-x86_64.tar.bz2\"\n",
        "else:\n",
        "  raise NotImplementedError"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AAX05CUVYyM9"
      },
      "source": [
        "!mkdir $blender_version\n",
        "if blender_version == \"blender2.79\":\n",
        "    !wget -O '{blender_version}.tar.xz' -nc $download_path\n",
        "    !tar xf '{blender_version}.tar.xz' -C ./$blender_version --strip-components=1\n",
        "else:\n",
        "  raise NotImplementedError"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Bw4p_etoY5LE"
      },
      "source": [
        "import os\n",
        "\n",
        "os.environ[\"LD_PRELOAD\"] = \"\"\n",
        "\n",
        "!apt update\n",
        "!apt remove libtcmalloc-minimal4\n",
        "!apt install libtcmalloc-minimal4\n",
        "os.environ[\"LD_PRELOAD\"] = \"/usr/lib/x86_64-linux-gnu/libtcmalloc_minimal.so.4.3.0\"\n",
        "\n",
        "!echo $LD_PRELOAD"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mo2mXUKbZPTN"
      },
      "source": [
        "!apt install libboost-all-dev\n",
        "!apt install libgl1-mesa-dev\n",
        "!apt install libglu1-mesa libsm-dev"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KG-U_Ym9aGm6"
      },
      "source": [
        "%cd /content/gdrive/MyDrive/clevr-dataset-gen/image_generation/ \n",
        "!echo $PWD >> blender2.79/2.79/python/lib/python3.5/site-packages/clevr.pth"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zpVm5OzsUJ8h"
      },
      "source": [
        "output_image_dir = YOUR_IMAGE_DIR\n",
        "output_scene_dir = YOUR_SCENE_DIR\n",
        "output_scene_file_path = YOUR_OVERALL_SCENE_FILE_PATH\n",
        "output_blend_dir = YOUR_OUTPUT_BLENDER_DIR[OPTIONAL]"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3Dc2q2_aDrit"
      },
      "source": [
        "!sudo ./$blender_version/blender \\\n",
        "--background --python render_images.py -- --num_images=9 --width=128 --height=128 \\\n",
        "--render_num_samples=64 --min_objects=1 --max_objects=5 --min_pixels_per_object=100 \\\n",
        "--output_image_dir=$output_image_dir \\\n",
        "--output_scene_dir=$output_scene_dir \\\n",
        "--output_scene_file=$output_scene_file_path \\\n",
        "--output_blend_dir=$output_blend_dir \\\n",
        "--min_dist=1 --margin=1 --use_gpu=1 &> /dev/null"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jMKmx3nNwY5g"
      },
      "source": [
        "import os\n",
        "import json\n",
        "import numpy as np\n",
        "from PIL import Image\n",
        "from matplotlib import pyplot as plt\n",
        "\n",
        "img_id = 0\n",
        "img_path = os.path.join(output_image_dir, 'CLEVR_new_{:06}.png'.format(img_id))\n",
        "im = Image.open(img_path).convert('RGB')\n",
        "im = np.array(im)\n",
        "\n",
        "scene_path = os.path.join(output_scene_dir, 'CLEVR_new_{:06}.json'.format(img_id))\n",
        "with open(scene_path, 'r') as f:\n",
        "  scene = json.load(f)\n",
        "\n",
        "obj_names = []\n",
        "for obj in scene['objects']:\n",
        "  name = ' '.join([obj['size'], obj['color'], obj['material'], obj['shape']])\n",
        "  obj_names.append(name)\n",
        "\n",
        "for relation, indices in scene['relationships'].items():\n",
        "  for i, js in enumerate(indices):\n",
        "    for j in js:\n",
        "      print(obj_names[j], relation, obj_names[i])\n",
        "\n",
        "if len(obj_names) == 1:\n",
        "  print(obj_names[0])\n",
        "\n",
        "plt.figure()\n",
        "plt.imshow(im)\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8NDj0rURX0c1"
      },
      "source": [
        "from PIL import Image\n",
        "import torchvision\n",
        "import torch\n",
        "import numpy as np\n",
        "from matplotlib import pyplot as plt\n",
        "\n",
        "images = []\n",
        "\n",
        "for i in range(9):\n",
        "  img_path = os.path.join(output_image_dir, 'CLEVR_new_{:06}.png'.format(i))\n",
        "  im = Image.open(img_path).convert('RGB')\n",
        "  im = np.array(im).transpose(2, 0, 1)\n",
        "  images.append(im)\n",
        "\n",
        "images = torch.from_numpy(np.array(images))\n",
        "grid = torchvision.utils.make_grid(images, nrow=3)\n",
        "\n",
        "plt.figure(figsize = (20,20))\n",
        "plt.imshow(grid.permute(1, 2, 0), interpolation='nearest')"
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}