notebooks/using_the_lib.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Import the image_embeddings lib"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "!pip install -U image_embeddings # don't run this line if running this directly in the repo"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import image_embeddings"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Download an example dataset\n",
    "\n",
    "For this notebook, I download a dataset from https://www.tensorflow.org/datasets/catalog/overview\n",
    "but any image can be used.\n",
    "For example :\n",
    "* tf_flowers\n",
    "* stanford_dogs\n",
    "* cats_vs_dogs\n",
    "* horses_or_humans\n",
    "* imagewang"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Let's define some paths where to save images, tfrecords and embeddings\n",
    "from pathlib import Path\n",
    "home = str(Path.home())\n",
    "dataset = \"tf_flowers\"\n",
    "path_images = f\"{home}/{dataset}/images\"\n",
    "path_tfrecords = f\"{home}/{dataset}/tfrecords\"\n",
    "path_embeddings = f\"{home}/{dataset}/embeddings\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:absl:Load pre-computed DatasetInfo (eg: splits, num examples,...) from GCS: tf_flowers/3.0.1\n",
      "INFO:absl:Load dataset info from /tmp/tmpzi_nv2hetfds\n",
      "INFO:absl:Generating dataset tf_flowers (/home/rom1504/tensorflow_datasets/tf_flowers/3.0.1)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1mDownloading and preparing dataset tf_flowers/3.0.1 (download: 218.21 MiB, generated: 221.83 MiB, total: 440.05 MiB) to /home/rom1504/tensorflow_datasets/tf_flowers/3.0.1...\u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:absl:Dataset tf_flowers is hosted on GCS. It will automatically be downloaded to your\n",
      "local data directory. If you'd instead prefer to read directly from our public\n",
      "GCS bucket (recommended if you're running on GCP), you can instead pass\n",
      "`try_gcs=True` to `tfds.load` or set `data_dir=gs://tfds-data/datasets`.\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "754190ac3c84421282fbd7dbf7b37441",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Dl Completed...', max=5.0, style=ProgressStyle(descriptio…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:absl:Load dataset info from /home/rom1504/tensorflow_datasets/tf_flowers/3.0.1.incompleteJG0WW4\n",
      "INFO:absl:Constructing tf.data.Dataset for split train, from /home/rom1504/tensorflow_datasets/tf_flowers/3.0.1\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "\u001b[1mDataset tf_flowers downloaded and prepared to /home/rom1504/tensorflow_datasets/tf_flowers/3.0.1. Subsequent calls will reuse this data.\u001b[0m\n"
     ]
    }
   ],
   "source": [
    "image_embeddings.downloader.save_examples_to_folder(output_folder=path_images, images_count=1000, dataset=dataset)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Transform image to tf records\n",
    "\n",
    "Tf record is an efficient format to store image, it's better to use than raw image file for inference"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shard 0 saved after 0s\n",
      "Shard 1 saved after 0s\n",
      "Shard 2 saved after 0s\n",
      "Shard 3 saved after 0s\n",
      "Shard 4 saved after 0s\n",
      "Shard 5 saved after 1s\n",
      "Shard 6 saved after 1s\n",
      "Shard 7 saved after 1s\n",
      "Shard 8 saved after 1s\n",
      "Shard 9 saved after 1s\n"
     ]
    }
   ],
   "source": [
    "image_embeddings.inference.write_tfrecord(image_folder=path_images,\n",
    "                                          output_folder=path_tfrecords,\n",
    "                                          num_shards=10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Build embeddings\n",
    "\n",
    "Here, efficientnet is used, but the code is particularly simple, and any other model could be used\n",
    "The input is tfrecords and the output is embeddings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1/1 [==============================] - 0s 14ms/step\n",
      "\n",
      "Shard 0 done after 6s\n",
      "Shard 0 saved after 6s\n",
      "1/1 [==============================] - 0s 7ms/step\n",
      "\n",
      "Shard 1 done after 11s\n",
      "Shard 1 saved after 11s\n",
      "1/1 [==============================] - 0s 5ms/step\n",
      "\n",
      "Shard 2 done after 16s\n",
      "Shard 2 saved after 16s\n",
      "1/1 [==============================] - 0s 5ms/step\n",
      "\n",
      "Shard 3 done after 21s\n",
      "Shard 3 saved after 21s\n",
      "1/1 [==============================] - 0s 9ms/step\n",
      "\n",
      "Shard 4 done after 26s\n",
      "Shard 4 saved after 26s\n",
      "1/1 [==============================] - 0s 5ms/step\n",
      "\n",
      "Shard 5 done after 31s\n",
      "Shard 5 saved after 31s\n",
      "1/1 [==============================] - 0s 6ms/step\n",
      "\n",
      "Shard 6 done after 36s\n",
      "Shard 6 saved after 36s\n",
      "1/1 [==============================] - 0s 7ms/step\n",
      "\n",
      "Shard 7 done after 41s\n",
      "Shard 7 saved after 41s\n",
      "1/1 [==============================] - 0s 7ms/step\n",
      "\n",
      "Shard 8 done after 45s\n",
      "Shard 8 saved after 46s\n",
      "1/1 [==============================] - 0s 5ms/step\n",
      "\n",
      "Shard 9 done after 50s\n",
      "Shard 9 saved after 50s\n"
     ]
    }
   ],
   "source": [
    "image_embeddings.inference.run_inference(tfrecords_folder=path_tfrecords,\n",
    "                                         output_folder=path_embeddings,\n",
    "                                         batch_size=1000)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Read the embeddings and build an index with it\n",
    "\n",
    "The knn index is built using https://github.com/facebookresearch/faiss which makes it possible\n",
    "to search embeddings in log(N) with lot of options to reduce memory footprint"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "[id_to_name, name_to_id, embeddings] = image_embeddings.knn.read_embeddings(path_embeddings)\n",
    "index = image_embeddings.knn.build_index(embeddings)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Search in the index\n",
    "\n",
    "Let's pick a random product by id, retrieve its embedding and search in the index\n",
    "\n",
    "Then let's display the closest products"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "image_dandelion_443\n"
     ]
    },
    {
     "data": {
      "image/jpeg": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "bf080f250d354dc0b119a42519eaa239",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(VBox(children=(Label(value='1.00 image_dandelion_443'), Image(value=b'\\xff\\xd8\\xff\\xe0\\x00\\x10J…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "p=225\n",
    "print(id_to_name[p])\n",
    "image_embeddings.knn.display_picture(path_images, id_to_name[p])\n",
    "results = image_embeddings.knn.search(index, id_to_name, embeddings[p])\n",
    "image_embeddings.knn.display_results(path_images, results)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Combination of images ?\n",
    "\n",
    "Any vector in the same space can be used as query\n",
    "For example I could have 2 image and want to find some example that are closeby to the 2, \n",
    "Let's just average them and see that happens !"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/jpeg": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/jpeg": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4045dd3c7d234b32a1be8284fbe20f2b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(VBox(children=(Label(value='134.38 image_dandelion_248'), Image(value=b'\\xff\\xd8\\xff\\xe0\\x00\\x1…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "p1 = 19\n",
    "p2 = 67\n",
    "image1 = id_to_name[p1]\n",
    "image2 = id_to_name[p2]\n",
    "image_embeddings.knn.display_picture(path_images, image1)\n",
    "image_embeddings.knn.display_picture(path_images, image2)\n",
    "results = image_embeddings.knn.search(index, id_to_name, (embeddings[p1] + embeddings[p2])/2, 7)\n",
    "image_embeddings.knn.display_results(path_images, results)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We get mostly one of the picture. One thing that can be done to improve this is to\n",
    "normalize the embeddings to get a better mix, let's try it !"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "\n",
    "def normalized(a, axis=-1, order=2):\n",
    "    l2 = np.atleast_1d(np.linalg.norm(a, order, axis))\n",
    "    l2[l2==0] = 1\n",
    "    return a / np.expand_dims(l2, axis)\n",
    "normalized_embeddings = normalized(embeddings, 1)\n",
    "index_normalized = image_embeddings.knn.build_index(normalized_embeddings)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/jpeg": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/jpeg": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "120642bd44994b69a716f8d420fe069f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(VBox(children=(Label(value='0.61 image_sunflowers_378'), Image(value=b'\\xff\\xd8\\xff\\xe0\\x00\\x10…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "p1 = 19\n",
    "p2 = 67\n",
    "image1 = id_to_name[p1]\n",
    "image2 = id_to_name[p2]\n",
    "image_embeddings.knn.display_picture(path_images, image1)\n",
    "image_embeddings.knn.display_picture(path_images, image2)\n",
    "results = image_embeddings.knn.search(index_normalized, id_to_name, (normalized_embeddings[p1] + normalized_embeddings[p2])/2, 7)\n",
    "image_embeddings.knn.display_results(path_images, results)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As expected we see a mixture of images that looks like both"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Exporting the embeddings to numpy\n",
    "For easy access to the embeddings in other languages, we provide a function to export them to numpy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from image_embeddings.knn import embeddings_to_numpy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "path_embeddings_numpy = f\"{home}/{dataset}/embeddings_numpy\"\n",
    "embeddings_to_numpy(path_embeddings, path_embeddings_numpy)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Next ?\n",
    "\n",
    "What else can you build with this ?\n",
    "\n",
    "An obvious example is retrieval (looking for closeby pictures) but recommendation systems or photo managers can also work very well using image embeddings !"
   ]
  },
  {
   "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": "3.7.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}