examples/step4a_initialize_annotations.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<a href=\"https://colab.research.google.com/github/jgraving/deepposekit/blob/master/examples/step4a_initialize_annotations.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# DeepPoseKit Step 4a - Initialize annotations\n",
    "\n",
    "This is step 4a of the example notebooks for using DeepPoseKit. This notebook shows you how to use your trained model to initialize the keypoint labels for the unannotated images in your annotation set.\n",
    "\n",
    "**NOTE**: If you run into problems, you can help us improve DeepPoseKit by [opening an issue](https://github.com/jgraving/deepposekit/issues/new) or [submitting a pull request](https://help.github.com/en/articles/creating-a-pull-request-from-a-fork)\n",
    "\n",
    "**If you're using Colab**: make sure to go to the “Runtime” dropdown menu, select “Change runtime type” and select `GPU` in the \"Hardware accelerator\" drop-down menu\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "If you haven't already installed DeepPoseKit you can run the next cell"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "!{sys.executable} -m pip install -U deepposekit"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import cv2\n",
    "import glob\n",
    "\n",
    "from deepposekit.models import load_model\n",
    "from deepposekit.io import DataGenerator, ImageGenerator\n",
    "\n",
    "from os.path import expanduser\n",
    "\n",
    "try:\n",
    "    import google.colab\n",
    "    IN_COLAB = True\n",
    "except:\n",
    "    IN_COLAB = False\n",
    "\n",
    "HOME = expanduser(\"~\") if not IN_COLAB else '.'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Use the next cell to download the example data into your home directory"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "!git clone https://github.com/jgraving/deepposekit-data {HOME + '/deepposekit-data'}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Find the saved models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['/Users/jake/deepposekit-data/datasets/fly/best_model_densenet.h5']"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "models = sorted(glob.glob(HOME + '/deepposekit-data/datasets/fly/*model*.h5'))\n",
    "models"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Find the annotation sets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['/Users/jake/deepposekit-data/datasets/fly/annotation_data_release.h5',\n",
       " '/Users/jake/deepposekit-data/datasets/fly/example_annotation_set.h5',\n",
       " '/Users/jake/deepposekit-data/datasets/fly/example_annotation_set1.h5']"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "annotations = sorted(glob.glob(HOME + '/deepposekit-data/datasets/fly/*annotation*.h5'))\n",
    "annotations"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Load the trained model\n",
    "\n",
    "This loads the trained model into memory for making predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING: Logging before flag parsing goes to stderr.\n",
      "W0929 12:52:50.678194 4792473024 deprecation.py:506] From /Users/jake/Library/Python/3.7/lib/python/site-packages/tensorflow/python/ops/init_ops.py:97: calling VarianceScaling.__init__ (from tensorflow.python.ops.init_ops) with dtype is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Call initializer instance with the dtype argument instead of passing it to the constructor\n",
      "W0929 12:52:50.679184 4792473024 deprecation.py:506] From /Users/jake/Library/Python/3.7/lib/python/site-packages/tensorflow/python/ops/init_ops.py:97: calling Zeros.__init__ (from tensorflow.python.ops.init_ops) with dtype is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Call initializer instance with the dtype argument instead of passing it to the constructor\n",
      "W0929 12:52:50.693815 4792473024 deprecation.py:506] From /Users/jake/Library/Python/3.7/lib/python/site-packages/tensorflow/python/ops/init_ops.py:97: calling GlorotUniform.__init__ (from tensorflow.python.ops.init_ops) with dtype is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Call initializer instance with the dtype argument instead of passing it to the constructor\n",
      "W0929 12:52:50.696883 4792473024 deprecation.py:506] From /Users/jake/Library/Python/3.7/lib/python/site-packages/tensorflow/python/ops/init_ops.py:97: calling Ones.__init__ (from tensorflow.python.ops.init_ops) with dtype is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Call initializer instance with the dtype argument instead of passing it to the constructor\n",
      "W0929 12:52:51.096096 4792473024 deprecation.py:323] From /Users/jake/Library/Python/3.7/lib/python/site-packages/tensorflow/python/keras/backend.py:4075: add_dispatch_support.<locals>.wrapper (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use tf.where in 2.0, which has the same broadcast rule as np.where\n"
     ]
    }
   ],
   "source": [
    "model = load_model(HOME + '/deepposekit-data/datasets/fly/best_model_densenet.h5')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Initialize unannotated data\n",
    "\n",
    "This creates a `DataGenerator` instance for loading the data, which is then wrapped in an `ImageGenerator`. `ImageGenerator` is then passed to the model for initializing the unannotated keypoints with predictions from a previously-trained model the image data. See the doc string for details:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "DataGenerator?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "ImageGenerator?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's initialize our `example_annotation_set.h5` from Step 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_generator = DataGenerator(HOME + '/deepposekit-data/datasets/fly/example_annotation_set.h5', mode='unannotated')\n",
    "image_generator = ImageGenerator(data_generator)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This passes the data generator to the model to get the predicted coordinates"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "100/100 [==============================] - 3s 31ms/sample\n"
     ]
    }
   ],
   "source": [
    "predictions = model.predict(image_generator, verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(100, 32, 3)"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predictions.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This saves the predicted coordinates as initial keypoint locations for the unannotated data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_generator[:] = predictions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Indexing the generator, e.g. `data_generator[0]` returns an image-keypoints pair, which you can then visualize. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "image, keypoints = data_generator[0]\n",
    "\n",
    "plt.figure(figsize=(5,5))\n",
    "image = image[0] if image.shape[-1] is 3 else image[0, ..., 0]\n",
    "cmap = None if image.shape[-1] is 3 else 'gray'\n",
    "plt.imshow(image, cmap=cmap, interpolation='none')\n",
    "for idx, jdx in enumerate(data_generator.graph):\n",
    "    if jdx > -1:\n",
    "        plt.plot(\n",
    "            [keypoints[0, idx, 0], keypoints[0, jdx, 0]],\n",
    "            [keypoints[0, idx, 1], keypoints[0, jdx, 1]],\n",
    "            'r-'\n",
    "        )\n",
    "plt.scatter(keypoints[0, :, 0], keypoints[0, :, 1], c=np.arange(data_generator.keypoints_shape[0]), s=50, cmap=plt.cm.hsv, zorder=3)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}