The SageMaker attached notebook that allows building an MXNet model that counts shapes in an image

notebook.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# You can modify these ones\n",
    "dataset_size = 12000\n",
    "img_size = 128\n",
    "shape_size = 8\n",
    "max_rows = 14\n",
    "max_cols = 14\n",
    "target_shape = \"circle\"\n",
    "\n",
    "# Do not modify these one\n",
    "shapes = [\"circle\", \"triangle\", \"square\"]\n",
    "img_width = img_height = img_size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "from random import randint, seed\n",
    "\n",
    "# Make sure the required folder structure is in place\n",
    "rmtree(\"data/raw\", True)\n",
    "os.makedirs(\"data/raw\")\n",
    "    \n",
    "# Make sure data generation is reproducible\n",
    "seed(42)\n",
    "\n",
    "with open(\"data/dataset.csv\", \"w+\") as dataset:\n",
    "    dataset.write(\"img_path,target\\n\")\n",
    "    for i in range(1, dataset_size + 1):\n",
    "        target_number_of_shapes = 0\n",
    "\n",
    "        filename = str(i).zfill(6) + \".shape\"\n",
    "\n",
    "        # Generate Shaper DSL source code\n",
    "        file_content = \"img_dim:{},shp_dim:{}>>>\".format(img_size, shape_size)\n",
    "        for x in range(0, randint(1, max_rows)):\n",
    "            for y in range(0, randint(1, max_cols)):\n",
    "                shape_idx = randint(0, len(shapes) - 1)\n",
    "                shape = shapes[shape_idx]\n",
    "                if shape == target_shape:\n",
    "                    target_number_of_shapes += 1\n",
    "                file_content = file_content + shape + \",\"\n",
    "            file_content = file_content[:-1]\n",
    "            file_content += \"|\"\n",
    "        file_content = file_content[:-1]\n",
    "        file_content += \"<<<\"\n",
    "\n",
    "        with open(\"./data/raw/{}\".format(filename), \"w\") as shape_file:\n",
    "            shape_file.write(file_content)\n",
    "\n",
    "        dataset.write(\"./raw/{}.png,{:d}\\n\".format(filename, target_number_of_shapes))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import subprocess\n",
    "\n",
    "subprocess.call(\"\"\"\n",
    "    java -cp bin/shaper-all.jar com.cosminsanda.shaper.compiler.Shaper2Image \\\n",
    "    --source-dir data/raw\"\"\", shell=True);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from pandas import read_csv\n",
    "import numpy as np\n",
    "\n",
    "df = read_csv(\"data/dataset.csv\")\n",
    "\n",
    "train = df.sample(frac=.8333, random_state=42)\n",
    "validation = df.loc[~df.index.isin(train.index), :] \\\n",
    "    .sample(frac=.5, random_state=42)\n",
    "test = df.loc[np.logical_not(\n",
    "                np.logical_xor(\n",
    "                    ~df.index.isin(train.index),\n",
    "                    ~df.index.isin(validation.index))), :]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test.size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import cv2\n",
    "import mxnet as mx\n",
    "\n",
    "def transform(row):\n",
    "    img = cv2.imread(\"./data/{}\".format(row[\"img_path\"]))\n",
    "    img = mx.nd.array(img)\n",
    "    img = img.astype(np.float32)\n",
    "    img = mx.nd.transpose(img, (2, 0, 1))\n",
    "    img = img / 255\n",
    "    label = np.float32(row[\"target\"])\n",
    "    return img, label"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_nd = [transform(row) for _, row in train.iterrows()]\n",
    "validation_nd = [transform(row) for _, row in validation.iterrows()]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "from pickle import dump\n",
    "\n",
    "def save_to_disk(data, type):\n",
    "    os.makedirs(\"data/pickles/{}\".format(type))\n",
    "    with open(\"data/pickles/{}/data.p\".format(type), \"wb\") as out:\n",
    "        dump(data, out)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "save_to_disk(train_nd, \"train\")\n",
    "save_to_disk(validation_nd, \"validation\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sagemaker\n",
    "\n",
    "sagemaker_session = sagemaker.Session()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "inputs = sagemaker_session.upload_data(path=\"data/pickles\", bucket=\"adt-adhoc\", key_prefix=\"cosmin/sagemaker/demo\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from shutil import rmtree, copy2\n",
    "\n",
    "rmtree(\"./test\", True)\n",
    "os.makedirs(\"./test\")\n",
    "for _, row in test.iterrows():\n",
    "    os.makedirs(\"test/{}\".format(row[\"target\"]), exist_ok=True)\n",
    "    copy2(\"./data/{}\".format(row[\"img_path\"]), \"./test/{}\".format(row[\"target\"]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "rmtree(\"data\", True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "estimator = sagemaker.mxnet.MXNet(\"object-counting-sagemaker-script.py\", \n",
    "          role=sagemaker.get_execution_role(), \n",
    "          train_instance_count=1, \n",
    "          train_instance_type=\"ml.p2.xlarge\",\n",
    "          hyperparameters={\"epochs\": 5},\n",
    "          py_version=\"py3\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "estimator.fit(inputs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "predictor = estimator.deploy(1, \"ml.m4.xlarge\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import boto3\n",
    "\n",
    "sagemaker_runtime_client = boto3.client(\"sagemaker-runtime\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import base64\n",
    "import json\n",
    "\n",
    "for directory in os.listdir(\"./test\"):\n",
    "    batch = []\n",
    "    for file in os.listdir(\"./test/{}\".format(directory)):\n",
    "        with open(\"./test/{}/{}\".format(directory, file), \"rb\") as image_file:\n",
    "            batch.append(base64.b64encode(image_file.read()).decode(\"utf-8\"))\n",
    "    binary_json = json.dumps(batch).encode(\"utf-8\")\n",
    "    response = sagemaker_runtime_client.invoke_endpoint(\n",
    "        EndpointName=predictor.endpoint,\n",
    "        Body=binary_json,\n",
    "        ContentType=\"application/json\",\n",
    "        Accept=\"application/json\"\n",
    "    )[\"Body\"].read()\n",
    "    individual_predictions = json.loads(response, encoding=\"utf-8\")\n",
    "    total = 0\n",
    "    detected = 0\n",
    "    for prediction in individual_predictions:\n",
    "        total += 1 \n",
    "        if int(prediction) == int(directory):\n",
    "            detected += 1\n",
    "    print(\"\"\"Images with {} circles:\n",
    "        Total: {}\n",
    "        Detected: {}\n",
    "        Accuracy: {:0.2f}\n",
    "    \"\"\".format(directory, str(total), str(detected), detected/total))"
   ]
  }
 ],
 "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}