Sample_S15EVA4_WFlow.ipynb

sample_s15eva4_wflow.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Sample_S15EVA4_WFlow.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyOhtkWktXievTNBwkI8CWi+",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/gist/rajy4683/5883b9d630a96f05f3fe72e6fc20e085/sample_s15eva4_wflow.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8TS3SyURpXEy",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "\"\"\"\n",
        "    Add-on functions to DenseDepth (https://github.com/ialhashim/DenseDepth.git)\n",
        "\"\"\"\n",
        "\"\"\"\n",
        "    \n",
        "    Helper function to load images from a ZipFile\n",
        "\n",
        "\"\"\"\n",
        "def load_execute_names(model, zipfile_orig, name_list,output_dir=\"/content/test\"):\n",
        "    image_list = [Image.open(BytesIO((zipfile_orig.read(name)))) for name in name_list ]   \n",
        "    inputs = load_images_local(image_list)\n",
        "    print('Loaded ({0}) images of size {1}.'.format(inputs.shape[0], inputs.shape[1:]))\n",
        "    outputs = predict(model, inputs)\n",
        "    converted_img = display_images_new(outputs.copy(),name_list,output_dir=output_dir)\n",
        "    print(\"Received output shape:{} converted:{}\".format(outputs.shape, converted_img.shape))\n",
        "\n",
        "\"\"\"\n",
        "    This function will save the predicted output array as JPEG image under '/content/test'\n",
        "    Inputs\n",
        "        outputs: output array \n",
        "        input_names_list: list of the filenames of the images\n",
        "    Output:\n",
        "        Colormapped JPEG files of depth images generated under '/content/test'\n",
        "    Can be called after the 'predict' function.\n",
        "    By default it uses the colormap as 'plasma'\n",
        "\"\"\"\n",
        "\n",
        "def display_images_new(outputs, input_names_list,output_dir='/content/test/' ):\n",
        "    is_colormap=True \n",
        "    is_rescale=True\n",
        "    plasma = plt.get_cmap('plasma')\n",
        "    shape = (outputs[0].shape[0], outputs[0].shape[1], 3)\n",
        "    \n",
        "    all_images = []\n",
        "\n",
        "    for i in range(outputs.shape[0]):\n",
        "        imgs = []        \n",
        "        if is_colormap:\n",
        "            rescaled = outputs[i][:,:,0]\n",
        "            if is_rescale:\n",
        "                rescaled = rescaled - np.min(rescaled)\n",
        "                rescaled = rescaled / np.max(rescaled)\n",
        "            imgs.append(plasma(rescaled)[:,:,:3])\n",
        "        else:\n",
        "            imgs.append(to_multichannel(outputs[i]))\n",
        "\n",
        "        img_set = np.hstack(imgs)\n",
        "        #print(\"Shape:{}\".format(np.array(img_set).shape))\n",
        "        file_name = os.path.join(output_dir,\"depth_\"+os.path.basename(input_names_list[i]))\n",
        "        Image.fromarray(np.uint8(np.array(img_set)*255)).save(file_name, 'JPEG',quality=100)\n",
        "        all_images.append(img_set)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Swn6g5eQ5nqG",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "########## Creating defaut output_dir ###########\n",
        "!mkdir /content/test"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fzrB01Pf5tUc",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "#### Sample workflow ######\n",
        "\n",
        "from DenseDepth import utils    #### load_execute_names\n",
        "from keras.models import load_model\n",
        "from layers import BilinearUpSampling2D\n",
        "\n",
        "os.environ['TF_CPP_MIN_LOG_LEVEL'] = '5'\n",
        "\n",
        "\n",
        "input_zip = ZipFile(\"/content/drive/My Drive/EVA4/base_images.zip\") ### You can directly read the file from Google-Drive path or from colab. Could't go beyond 300-400 files at a time\n",
        "file_list = [ name for name in depth_zip.namelist()][:200]\n",
        "\n",
        "custom_objects = {'BilinearUpSampling2D': BilinearUpSampling2D, 'depth_loss_function': None}\n",
        "print('Loading model...')\n",
        "# Load model into GPU / CPU\n",
        "model = load_model('/content/nyu.h5', custom_objects=custom_objects, compile=False)\n",
        "utils.load_execute_names(model, input_zip, file_list)"
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}