thermo/convert_optimize_quantize.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "2370b7f2",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'\n",
    "from pathlib import Path\n",
    "\n",
    "import yaml\n",
    "import numpy as np\n",
    "import seaborn as sns\n",
    "import tensorflow as tf\n",
    "import matplotlib.pyplot as plt\n",
    "from tqdm import tqdm\n",
    "\n",
    "from models.unet import UNet\n",
    "from utils.model_utils import check_model_prediction, evaluate, tflite_inference\n",
    "from metrics import CountAccuracy, CountMAE, CountMeanRelativeAbsoluteError\n",
    "from data_generator.thermal_data_generator import ThermalDataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "1662e661",
   "metadata": {},
   "outputs": [],
   "source": [
    "MODEL_NAME = 'model'\n",
    "MODELS_DIR = './data/model'\n",
    "MODEL_DTYPE = 'fp32' # uint8, int8, fp16, fp32"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "0bcb4cfd",
   "metadata": {},
   "outputs": [],
   "source": [
    "with open('./configs/config.yaml', 'r') as stream:\n",
    "    config = yaml.safe_load(stream)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2bdc0dcf",
   "metadata": {},
   "source": [
    "## Test data generator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "b7700921",
   "metadata": {},
   "outputs": [],
   "source": [
    "test_data_gen = ThermalDataset(\n",
    "    data_path=Path('./data'), \n",
    "    sequences_names=config['dataset'][\"test_dirs\"],\n",
    "    person_point_weight=config['dataset']['sum_of_values_for_one_person'], \n",
    "    batch_size=1\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "19fb04ae",
   "metadata": {},
   "source": [
    "## Test single frame"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "1bc79f10",
   "metadata": {},
   "outputs": [],
   "source": [
    "input_frame = test_data_gen[0][0]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8673280c",
   "metadata": {},
   "source": [
    "# TF Keras model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "4215ed1c",
   "metadata": {},
   "outputs": [],
   "source": [
    "model = tf.keras.models.load_model(f'{MODELS_DIR}/{MODEL_NAME}.h5', compile=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "838256db",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4.66686571932814\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "output_frame = model.predict(input_frame, verbose=0)\n",
    "\n",
    "print(np.sum(output_frame) / config['dataset'][\"sum_of_values_for_one_person\"])\n",
    "\n",
    "fig, ax = plt.subplots(1,2, figsize=(10,4));\n",
    "ax[0].imshow(input_frame.reshape(config['dataset'][\"IR_camera_resolution\"]));\n",
    "ax[1].imshow(output_frame[0][..., 0]);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "28fcdc55",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of tested frames: 2347\n",
      "Average inference time: 0.031761\n",
      "Average FPS: 31.4851\n",
      "Model Accuracy = 0.9795483596080102\n",
      "Model F1 score = 0.9798091302816225\n",
      "Predicted:\n",
      "   165 frames with 0 persons\n",
      "   83 frames with 1 persons\n",
      "   220 frames with 2 persons\n",
      "   333 frames with 3 persons\n",
      "   1229 frames with 4 persons\n",
      "   316 frames with 5 persons\n",
      "   1 frames with 6 persons\n",
      "Predicted correctly:\n",
      "   162 frames with 0 persons\n",
      "   78 frames with 1 persons\n",
      "   205 frames with 2 persons\n",
      "   328 frames with 3 persons\n",
      "   1220 frames with 4 persons\n",
      "   306 frames with 5 persons\n",
      "mae: 0.10517487009357382\n",
      "mse: 0.03238200663388619\n",
      "mae_rounded: 0.020877716233489562\n",
      "mse_rounded: 0.021729867916489135\n",
      "mrae: 0.007832694219571086\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "test_acc, test_f1, test_cm = evaluate(f'{MODELS_DIR}/{MODEL_NAME}.h5', 'keras', test_data_gen, config)\n",
    "\n",
    "plt.figure(figsize=(5,4));\n",
    "sns.heatmap(\n",
    "    test_cm, annot=True, cmap=\"Blues\", fmt='0.0f', \n",
    "    xticklabels=np.arange(config['dataset'][\"max_people_count\"]+1), \n",
    "    yticklabels=np.arange(config['dataset'][\"max_people_count\"]+1)\n",
    ");\n",
    "plt.xlabel('Labels');\n",
    "plt.ylabel('Predictions');"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b3e5c735",
   "metadata": {},
   "source": [
    "## TF Lite conversion"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "354eec2d",
   "metadata": {},
   "outputs": [],
   "source": [
    "def representative_dataset():\n",
    "    repr_data_gen = ThermalDataset(\n",
    "        data_path=Path('./data'), \n",
    "        sequences_names=config['dataset'][\"training_dirs\"],\n",
    "        person_point_weight=config['dataset']['sum_of_values_for_one_person'], \n",
    "        batch_size=1\n",
    "    )\n",
    "    \n",
    "    frames_num = repr_data_gen.__len__()\n",
    "    \n",
    "    for i in tqdm(range(frames_num)):\n",
    "        data = repr_data_gen[i][0]\n",
    "        data = data.astype(np.float32)\n",
    "        yield [data]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "7a6dceb3",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: /tmp/tmpvjflfgy4/assets\n"
     ]
    }
   ],
   "source": [
    "# Convert the model.\n",
    "converter = tf.lite.TFLiteConverter.from_keras_model(model)\n",
    "if MODEL_DTYPE == 'fp16':\n",
    "    converter.optimizations = [tf.lite.Optimize.DEFAULT]\n",
    "    converter.target_spec.supported_types = [tf.float16]\n",
    "elif MODEL_DTYPE == 'int8':\n",
    "    converter.optimizations = [tf.lite.Optimize.DEFAULT]\n",
    "    converter.representative_dataset = representative_dataset\n",
    "    converter.target_spec.supported_ops = [tf.lite.OpsSet.TFLITE_BUILTINS_INT8]\n",
    "    converter.target_spec.supported_types = [tf.int8]\n",
    "    converter.inference_input_type = tf.int8\n",
    "    converter.inference_output_type = tf.int8\n",
    "elif MODEL_DTYPE == 'uint8':\n",
    "    converter.optimizations = [tf.lite.Optimize.DEFAULT]\n",
    "    converter.representative_dataset = representative_dataset\n",
    "    converter.target_spec.supported_ops = [tf.lite.OpsSet.TFLITE_BUILTINS_INT8]\n",
    "    converter.target_spec.supported_types = [tf.int8]\n",
    "    converter.inference_input_type = tf.uint8\n",
    "    converter.inference_output_type = tf.uint8\n",
    "tflite_model = converter.convert()\n",
    "\n",
    "# Save the model.\n",
    "with open(f'{MODELS_DIR}/{MODEL_NAME}_{MODEL_DTYPE}.tflite', 'wb') as f:\n",
    "    f.write(tflite_model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "0c5e4c1c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4.66686571932814\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "output_frame = tflite_inference(f'{MODELS_DIR}/{MODEL_NAME}_{MODEL_DTYPE}.tflite', input_frame)\n",
    "\n",
    "print(np.sum(output_frame[0]) / config['dataset'][\"sum_of_values_for_one_person\"])\n",
    "\n",
    "fig, ax = plt.subplots(1,2, figsize=(10,4));\n",
    "ax[0].imshow(input_frame[0,:,:,0]);\n",
    "ax[1].imshow(output_frame[0][0,:,:,0]);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "7e108ccc",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of tested frames: 2347\n",
      "Average inference time: 0.000773\n",
      "Average FPS: 1292.9767\n",
      "Model Accuracy = 0.9795483596080102\n",
      "Model F1 score = 0.9798091302816225\n",
      "Predicted:\n",
      "   165 frames with 0 persons\n",
      "   83 frames with 1 persons\n",
      "   220 frames with 2 persons\n",
      "   333 frames with 3 persons\n",
      "   1229 frames with 4 persons\n",
      "   316 frames with 5 persons\n",
      "   1 frames with 6 persons\n",
      "Predicted correctly:\n",
      "   162 frames with 0 persons\n",
      "   78 frames with 1 persons\n",
      "   205 frames with 2 persons\n",
      "   328 frames with 3 persons\n",
      "   1220 frames with 4 persons\n",
      "   306 frames with 5 persons\n",
      "mae: 0.10517487267117653\n",
      "mse: 0.032382006621022585\n",
      "mae_rounded: 0.020877716233489562\n",
      "mse_rounded: 0.021729867916489135\n",
      "mrae: 0.007832694219571086\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAUUAAAEHCAYAAAA55FQ6AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/MnkTPAAAACXBIWXMAAAsTAAALEwEAmpwYAAAvxElEQVR4nO3dd3wUdfrA8c+ThNBbgCwtKhg8C9hADkWpUqQ3RTgbIlF+WFEEFUFsqHfq2U4FAfEUuycoHBYEURTBgoCiJyJCEBKkl0CSzfP7YydxRUg268zuZvO8ec0r03ae76Q8fGe+8/2OqCrGGGMCEqJdAGOMiSWWFI0xJoglRWOMCWJJ0RhjglhSNMaYIJYUjTEmSFK0C3Ak67YeiOqzQg1rV4pmeGOiplISEs7nKp92dch/szlfPV5sDBGZDvQCslW1ubPu70BvIBf4ERimqjudbbcAwwE/cK2qvuOs7w48AiQCz6jqfSWVzWqKxhh3SELoU8meBbofsu49oLmqngz8D7gFQEROBC4ETnI+8y8RSRSRROAJ4DzgRGCIs2+xLCkaY9whEvpUAlVdDGw/ZN27qprvLC4FGjvzfYGXVPWgqv4ErAVaO9NaVV2nqrnAS86+xbKkaIxxh7s1xZJcDvzXmW8EbAzalumsO9L6YllSNMa4oxQ1RRHJEJHPg6aM0MPIbUA+8IIXpxGzDS3GmDImITHkXVV1CjCltCFE5DICDTCd9beBGzYBaUG7NXbWUcz6I7KaojHGHR5fPjstyTcDfVR1f9CmOcCFIlJRRJoAzYBlwHKgmYg0EZFkAo0xc0qKYzVFY4w7QmhACf1Q8iLQAagrIpnARAKtzRWB9yQQa6mqXqWq34jIK8C3BC6rR6mq3znO1cA7BB7Jma6q35QYO1aHDrPnFI2JjrCfUzzr1tCfU/zkXvcyqMuspmiMcYeLNcVosqRojHGHO4/aRF2ZPYuH7p3Ahb06cNXFA363fvZrsxgxtC9XXtSfaf96GIAvl3/KNZdfyMhLBnLN5Rey4ovPPC3bko8W06dnN3p178K0qaVuYCuzsS1+OY+fkBj6FMPKbE2xS4++9Bk4hH/cfVvRuq+/XMbSjxbxxLOvkpyczM4d2wCoUbMWdzzwKHXqprJ+3Q+MHz2S599835Ny+f1+7r3nTp6eOgOfz8fQwYPo0LETx6anexIvVmJbfItvNcUoa3FqS6rXqPG7dXP/8yoXXHQ5ycnJANSqXQeA9ONOoE7dVACObpLOwYMHyc3N9aRcq1etJC3taBqnpVEhOZnuPXqyaOECT2LFUmyLb/FJkNCnGOZZUhSR40VkrIg86kxjReQEr+IBbNr4M6tXfsn1I/7GmKsv5/s1q/+wz8eL3if9uBOKEqfbsrOyqN+gftFyqs9HVlaWJ7FiKbbFt/gR7ubnGU9KJyJjCXS+FgIPUS5z5l8UkXFexATw+/PZs3sXD095niv+7wYmTxhD8CNHP69by/Qn/8k1N9/uVRGMKb9cHBAimrxK2cOBM1T1PlV93pnuIzBqxfAjfSi4P+SLz00rddC69Xy0bd8ZEeEvJ7ZAJIFdO3cAsDU7i7tuvYGbxt9Nw0ZpJRwpfKk+H1s2bylazs7KwufzeRYvVmJbfIsfLw0tXiXFAqDhYdY3cLYdlqpOUdVWqtpqyCVHzJ1HdGa7jnz95XIAMjesJz8/j5q1arN3z24mjrmaYSOv46STTyv1cUvjpOYt2LBhPZmZG8nLzWX+vLm079jJ05ixENviW/x4uXz2qvX5emCBiPzAb0P3HAWkA1e7EeC+iWNZueJzdu/cyUX9u3Dx8JF07dmfhydP4KqLB5BUoQI33nYXIsJbr7/EL5s2MGvGFGbNCDymcM/DTxY1xLgpKSmJW26bwMiMKygo8NOv/0DS05u5HifWYlt8ix/rl8Wh8qybn4gkELhcLhy/bBOwvLBPYkmsm58x0RF2N7/uD4XezW/+6JjNoJ49p6iqBQRGxzXGlAdxUlMssw9vG2NiTIzfKwyVJUVjjDtivFU5VJYUjTHusJqiMcYEsXuKxhgTxGqKxhgTxGqKxhgTxGqKxhjzG0mwpGiMMUXELp+NMSZIfOTE2E2K0e57vOdAflTjV68Usz8aYw7LaorGGBPEkqIxxgRJsIYWY4wJEh8VRUuKxhh3xMvlc3zUd40xUSciIU8hHGu6iGSLyOqgdSki8p6I/OB8re2sF+eNoWtFZKWInB70mUud/X8QkUtDOQ9LisYYV7iZFIFnge6HrBsHLFDVZsACZxngPKCZM2UATzrlSQEmAn8l8BaAiYWJtDiWFI0xrnAzKarqYmD7Iav7AjOd+ZlAv6D1z2nAUqCWiDQAugHvqep2Vd0BvMcfE+0f2D1FY4wrJMHze4o+Vd3szG8BCt/f2ojfXpAHkOmsO9L6YllN0RjjitLUFIPf8e5MGaWJpYE37nnycjurKRpjXFGa1mdVnQJMKWWILBFpoKqbncvjbGf9JiAtaL/GzrpNQIdD1i8qKUjc1RQPHjzI0MGDOL9/H/r36cm/Hn/U85gb1v/EZUMGFE1d27XmlVnP8cP3a8i4dAiXDRnA8Isu4NvVKz0tx5bNmxl+2cX0792D/n168sK/Z5b8IZct+WgxfXp2o1f3LkybWtrfeYtfpuNLKabwzAEKW5AvBWYHrb/EaYVuA+xyLrPfAbqKSG2ngaWrs65YcVdTTE5O5pnpM6lStSp5eXlcdvFQzj6nHSefcqpnMY86pgnPvvgGAH6/n/7ndaRdx3O5/66JDMv4P85sew6ffryYfz36EI9PedazciQmJXLTzeM44cST2LdvLxeeP5A2Z7bl2PR0z2IG8/v93HvPnTw9dQY+n4+hgwfRoWMni19O4rv5nKKIvEiglldXRDIJtCLfB7wiIsOBn4ELnN3nAT2AtcB+YBiAqm4XkbuA5c5+d6rqoY03fxB3SVFEqFK1KgD5+fnk5+dHdETgL5YtpVHjNOo3aIgI7N+3F4C9e/dQt249T2PXq5dKvXqpAFStWo2mTZuSnZ0VsT+K1atWkpZ2NI3TAlcy3Xv0ZNHCBRa/nMR3Mymq6pAjbOp8mH0VGHWE40wHppcmdtwlRQj8jznk/AFs2LCBwUOGcvLJp0Qs9vvv/pdzu/UA4NqbxjF6VAZP/PMfFBQU8NSMFyJWjk2bMvluzRpaRPDcs7OyqN+gftFyqs/HqpXe3jKw+LETP176Pkf8LERkmNcxEhMTeeWN2bz7wYesXrWSH374n9chAcjLy2XJhwvpeG43AN589WWuvXEsb8xbwDWjxzL5ztsjUo79+/Zx4/XXMmbcrVSrVi0iMY2JwD3FiIhGap90pA3BzfRu3CSuUaMGZ7T+K598/NGfPlYoli75mOOOP5GUOnUB+O/bs2nfqQsAnbp0Y803qzwvQ15eHqOvv5YePXtzbpeunscLlurzsWXzlqLl7KwsfD5fMZ+w+PEU3+UeLVHjSVJ0+h8eblrFbw9c/oGqTlHVVqraaviIUj22VGT79u3s3r0bgAMHDrD00084pknTsI5VWu+/M49zu/coWq5bL5Wvvgjc4/1i+Wc0Tjva0/iqyh0TbqNp06ZccpnnFfI/OKl5CzZsWE9m5kbycnOZP28u7Tt2svjlJH68JEWv7in6CHSx2XHIegE+8SgmAL9uzWb8reMoKPBTUKB07dad9h06ehkSgJyc/Sz/7BPG3DqxaN3N4+/gkX/ch9+fT3JyRW4ef4enZfjqyy94e85smh13HBcM6AvANdeP5px27T2NWygpKYlbbpvAyIwrKCjw06//QNLTm0UktsWPfvxYT3ahkkDDjcsHFZkGzFDVjw+zbZaqDi3pGAfyvXlaPVT2OgJTXlVKCu+u31HXzAn5b3bDY31iNoN68penqsOL2VZiQjTGlD3xUlO06ogxxhWWFI0xJoglRWOMCRYfOdGSojHGHVZTNMaYIAneDzIbEZYUjTGusJqiMcYEiZOcaEnRGOMOqykaY0yQOMmJlhSNMe6whpY4V61idL81P2Xvi2r8JqlVoxrflD2WFI0xJohdPhtjTBBraDHGmCCWFI0xJkic5ERLisYYd1hDizHGBLHLZ2OMCRInOdGSojHGHfFSU4zGe5+NMXFIJPQptOPJDSLyjYisFpEXRaSSiDQRkc9EZK2IvCwiyc6+FZ3ltc72Y8I9D0uKxhhXuPneZxFpBFwLtFLV5kAicCFwP/CwqqYTeIVy4UvyhgM7nPUPO/uFJS4vn5d8tJj777uHAn8B/Qeez/ARGRGNf17XTlStWpWEhASSEhOZ9cobrh5/a/YWHpk8gZ07tiEIXXsNoPegoezZvYt/3DmO7C2/kFq/IWMm3k+16jVYteJzJo8fTWr9hgCceU4nBl/qzfck2t/7CeNvYfGHi0hJqcMbs9+OaGyI/vlHM74Hrc9JQGURyQOqAJuBTkDhG0FnAncATwJ9nXmA14DHRUQ0jHc4x11S9Pv93HvPnTw9dQY+n4+hgwfRoWMnjk1Pj2g5pk6fSe3aKZ4cOzExkWEjb+DY404gZ/8+brzyb5zaqg0L5s/h5NNbM3DoMF6fNYPXZ83g0iuvA+DEFqcyfvKjnpSnUCx87/v2G8CQoRdx2y1jIxazULTPP9rx3bylqKqbROQfwAYgB3gX+ALYqaqFL2XPBBo5842Ajc5n80VkF1AH+LW0sePu8nn1qpWkpR1N47Q0KiQn071HTxYtXBDtYrkqpU49jj3uBAAqV6lK46OasO3XbJZ98iEdu/UCoGO3Xny2ZFFEyxUL3/uWrc6gRs2aEY1ZKNrnH+34pbl8FpEMEfk8aMo45Fi1CdT+mgANgapA90ich2dJUUSOF5HOIlLtkPWenlh2Vhb1G9QvWk71+cjKyvIy5B+IwMiM4Qy5YACvvfqyp7GytvzCurXfc9wJzdm5fRspdeoBUDulLju3byva7/tvV3H98MHcOfZqNvz0oydliYXvfTRF+/yjHb80DS2qOkVVWwVNUw453LnAT6q6VVXzgDeAtkAtESm8wm0MbHLmNwFpgXJIElAT2EYYPLl8FpFrgVHAGmCaiFynqrOdzfcC872IGytmPPciPp+P7du2cdWIYTRp0pSWrc5wPU5Ozn7un3ATw0fdSJWqv/u/53c3tI9tdjxTXppL5cpV+Hzpx0y+fTRPPj/7cIc0JmwuP5KzAWgjIlUIXD53Bj4HFgKDgJeAS4HCX+Q5zvKnzvYPwrmfCN7VFEcALVW1H9ABuF1ErnO2HfE7F1ylnjb10P84QpPq87Fl85ai5eysLHw+X1jHCldhvJQ6dejYuQurV610PUZ+fh73T7iJ9uf24Mx2nQGolVKH7du2ArB921ZqOvc0q1StRuXKVQBo1eZs8vPz2b1rh+tlioXvfTRF+/yjHd/N1mdV/YxAg8mXwCoCuWoKMBYYLSJrCdwznOZ8ZBpQx1k/GhgX7nl4lRQTVHUvgKquJ5AYzxORhygmKQZXqcNtNTupeQs2bFhPZuZG8nJzmT9vLu07dgrrWOHI2b+fffv2Fs1/+skS0ps1czWGqvL4A3fS+Ogm9L3goqL1rc9qx8J3Ai2uC995m9ZntQdgx/ZfKfxP839rVqOqVK9Ry9UyQfS/99EW7fOPdvyEBAl5CoWqTlTV41W1uaperKoHVXWdqrZW1XRVPV9VDzr7HnCW053t68I9D69an7NE5FRVXQGgqntFpBcwHWjhUUwAkpKSuOW2CYzMuIKCAj/9+g8kPd3dpFScbdu2Mfq6UQDk+/2c16MXbc9u52qMNatXsOi9uRzdNJ3rr7gQgIuuuJoBQ4bx90ljeX/em9TzNWDMxMCjWp98+D7zZ79GYmIiyRUrctPtkz3pfRDt7z3A2JtG8/nyZezcuYMundoxctQ1DBh4fkRiR/v8ox0/Tjq0IGFedhd/UJHGQL6qbjnMtraquqSkYxzIx/2ClYIH35ZSWb/VXkdgoqNS0pGv5orT6dFPQ/6r+eDaM2M2hXpSU1TVzGK2lZgQjTFlT7zUFOPu4W1jTHQkxElWDKmhRUSOFZGKznwHEblWRGp5WjJjTJnidkNLtITa+vw64BeRdALN4mnALM9KZYwpcxIk9CmWhXr5XOD0J+wPPKaqj4nIV14WzBhTtsTLeIqhJsU8ERlC4Inx3s66Ct4UyRhTFsVJTgz58nkYcCZwj6r+JCJNgH97VyxjTFkjpfgXy0KqKarqtwQGfCxc/ok/MYijMSb+xPq9wlCFlBRFpC2BARyPdj4jgKpqU++KZowpS2K9VTlUod5TnAbcQGCQR793xTHGlFXx8pxiqElxl6r+19OSGGPKtDjJiSEnxYUi8ncCAz0eLFypql96UipjTJlT3h7J+avztVXQOiXwEpm4FO2f79H1qkQ1/rJ126Mav3VTb95vY7wT7b8Zt4Ta+tzR64IYY8q2xDjJiqH2fa4pIg8FvWTmQRGJztuBjDExyc2Rt6Mp1Ie3pwN7gAucaTcww6tCGWPKnvLW9/lYVR0YtDxJRFZ4UB5jTBkV6zXAUIVaU8wRkbMLF5yHuXO8KZIxpiwqzStOY1moNcWRwEznPqIA24HLvCqUMabsiZeaYqitzyuAU0SkhrO828tCGWPKnsRYv1kYomKToohcpKrPi8joQ9YDoKoPeVg2Y0wZEh8pseSaYuEr3aofZluU33dnjIkl5aLvs6o+7cy+f+hb+JzGlpi05KPF3H/fPRT4C+g/8HyGj8iI6/h3jL+VxYsXkZJSh9fefAuAJx57hA8/WIAkJJCSksKkeyaTmupzJV5e7kEeGDeS/Lw8/H4/Ldt2pO/fRjD1HxP5ee13JCYm0eS4E7ho1DiSkpLYv28v0x68g+1bs/D7/XQbMJS25/ZypSyHKm8/+1iKHyc5MbT3PovIl6p6eknr3BTue5/9fj99enbj6akz8Pl8DB08iPv+/hDHpqe7XURP4xeU4sXTX3y+nCpVqnD7reOKkuLevXupVq0aALOef451P/7I+ImTQj7m5z/tOOI2VeXggRwqVa5Cfn4+D4y9ksEjbmD/3t00b3kmAFP/MZHjTjqVDj0GMPeVZ8nZv49Bl41iz64djL9qMA8+N5ekCkcevD2cbn7x8rOPdvxw3/uc8eo3If/STjn/pJhNocU+kiMiZ4rIjUA9ERkdNN0BJEakhKW0etVK0tKOpnFaGhWSk+neoyeLFi6I6/gtW51BzZq/72BUmBABcnJyXG0ZFBEqVQ70zfbn5+PPz0dEaNHqrKIeC02ancCOX7OL9j+4fz+qyoGcHKpWr0FCovu/PuXxZx9L8ePlkZySnlNMBqoRuMyuHjTtBgYV90ERaS0iZzjzJzrJtMefL3LxsrOyqN+gftFyqs9HVlaW12FjJn6wxx95mO6dO/DfuW8z8uprS/5AKRT4/Uy69hJuvLgHJ5zWmqZ/OaloW35+PksXzueklm0A6NRzEJsz1zPm0t5MuuYiLhxxAwkJoT4iG7pof+/Le/zEBAl5CoWI1BKR10TkOxFZ41TSUkTkPRH5wfla29lXRORREVkrIitFJOyr2GJ/M1X1Q1WdBLRR1UlB00Oq+kMxJzMReBR4UkQmA48TaLQZJyK3hVtYUzpXX3cD8xcs4ryevXh51vOuHjshMZGJjz7HAzNms/5/37Lp5x+Lts168u80a34qx510KgDffPUZaU2a8feZbzHhkZnMeupBcvbvc7U8Jvo86Pv8CDBfVY8HTgHWAOOABaraDFjgLAOcBzRzpgzgyXDPI9T/rp8RkVqFCyJSW0TeKWb/QUBboB0wCuinqncB3YDBR/qQiGQUDjoxbeqUEIv2e6k+H1s2bylazs7Kwudzp4GhLMQ/nB69erPg/fc8OXaVatX5S4vTWf3FUgDmvDiNPbt2csHw64r2WfL+XE47qwMiQmrDNOrWb8iWzPWulyXa3/vyHj+hFFNJnI4i7QiM+o+q5qrqTqAvMNPZbSbQz5nvCzynAUuBWiLSINzzCEVdp0A4BdwBpBazf76q+lV1P/Bj4cPeqpoDFBzpQ6o6RVVbqWqrcFvNTmregg0b1pOZuZG83Fzmz5tL+46RG/Yx2vEL/fzz+qL5RR8s4JgmTVw79p5dO9i/dw8AuQcP8O2K5dRvfDQfvTOHb79cyogxk353eZxSz8d3X38OwO4d28nK/Jm6vkauladQtL/35T2+yzXFJsBWYIaIfCUiz4hIVcCnqpudfbYAhVm/EbAx6POZzrpSC7WbX4GIHKWqGwBE5GiKf04xV0SqOEmxZeFKJ/sfMSm6ISkpiVtum8DIjCsoKPDTr/9A0tObeRky6vHHjRnNF8uXs3PnDrp1bs9V/3cNH3/0IT+vX0+CCA0aNuS2CaG3PJdk1/ZtTP/nnRQUFKAFSquzO3FK67O5su/Z1Emtz+Qxgf/QTj+zPb2HDKfX4GHM+Ofd3HH131CFgZeNonrNWq6Vp1B5/NnHUvzSdGgRkQwCl7mFpqhq8OVhEnA6cI2qfiYij/DbpTIQeHOeiLj+vHSoj+R0B6YAHxJ4cP0cIENVD3sJLSIVVfXgYdbXBRqo6qqSYob7SE68KM0jOV4o7pGcSLCRt6Mn3Edybnzr+5B/aR/s/ZdiY4hIfWCpqh7jLJ9DICmmAx1UdbNzebxIVf8iIk878y86+39fuF9pzyOky2dVnU8ga78MvAS0PFJCdPb/Q0J01v8aSkI0xpQ9bo6nqKpbgI0i8hdnVWfgW2AOcKmz7lJgtjM/B7jEaYVuQ+Ble6VOiFBy3+fjVfW7oObtX5yvRzmX0/biKmMM4Mnzh9cAL4hIMrAOGEagIveKiAwHfiYw6DXAPKAHsBbY7+wblpLuKd4IjAAePMy2uH5xlTGmdNzu++yMztXqMJs6H2ZfJfCky59WUt/nEc5Xe3GVMaZY7j+OHx0lXT4PKG67qr7hbnGMMWVVrHffC1VJl8+9na+pwFnAB85yR+ATwJKiMQYoJ4PMquowABF5FzixsDXHaQp/1vPSGWPKjDjJiSE/vJ12SPN2FnCUB+UxxpRR5WKQ2SALnL7OLzrLg4H3vSmSMaYsipOcGPKLq64Wkf4EOmhDoEvOf7wrljGmrClvl88AXwJ7VPV9EakiItVVdY9XBTPGlC0SJ6+uCikpisgIAp23U4BjCYw+8RSHeYjSGFM+JcXJg4qh1hRHAa2BzwBU9QcRKW7oMPMnRfum9RlNojsgQ+0zro5q/G3LHotq/Gj//MPh5isvoinUpHhQVXMLT1pEkrBXnBpjgsTLPcVQK7wfisitQGUR6QK8CrzlXbGMMWVNeXlxVaGxBEbBXQVcSWBEivFeFcoYU/YkiIQ8xbISL59FJBH4xnl5zFTvi2SMKYsS46ShpcTTUFU/8L2IWA8WY8wRJSAhT7Es1IaW2sA3IrIMKHo3par28aRUxpgyJ8avikMWalK83dNSGGPKvHhpfS5pPMVKwFUEXhazCpimqvmRKJgxpmyJ9QaUUJVUU5wJ5AEfAecBJwLXFfsJY0y5FCc5scSkeKKqtgAQkWnAMu+L9Oct+Wgx9993DwX+AvoPPJ/hIzJK/lCcxI/2uZ/XtRNVq1YlISGBpMREZr3izjjET038G+e1a87W7Xtodf69ANx7fT96tGtObp6fnzJ/JWPi8+zam0Onvx7PXdf2IblCErl5+dz6zzf5cPn/ADjthDSmTLqYyhUr8M6Sb7jxgdf+VLnuGH8rixcvIiWlDq+9GXh0d9eunYy9cTS//LKJhg0b8cCDD1OjZs0/9w0IUTR//vEyyGxJrc95hTNl5bLZ7/dz7z138q+nnuE/c+Yyf97b/Lh2bbmIH+1zLzR1+kxeeX22awkR4N9vLaXvqCd+t27B0u9oef69tB48mR9+zmbM5V0B2LZzL4Ouf5ozLriXERP+zfS7Lyn6zKO3DmbUXbNo3ncSxx5Vj65tT/xT5erdrz9PPPX7J9VmPDOV1m3aMGfeO7Ru04YZ0yLzJFu0f/4JpZhiWUnlO0VEdjvTHuDkwnkR2R2JApbW6lUrSUs7msZpaVRITqZ7j54sWrigXMSP9rl7acmXP7J91/7frVuw9Dv8/gIAlq36iUa+WgB8/X0mm7fuAuDbHzdTqWIFkiskUb9uDapXrcSyVesBmPX2Mnp3OPlPlatlqzOoeUgtcNHCBfTu2w+A3n37sfCDyAw9Gu2fv4iEPMWyYpOiqiaqag1nqq6qSUHzNSJVyNLIzsqifoP6RcupPh9ZWVnlIn60zx0C95VGZgxnyAUDeO3VlyMW95K+Z/LOkm//sL7/uaey4ruN5Obl0zC1FpuydxZt25S1k4aptVwvy7Zt26hXLzBeSt269di2bZvrMQ4n2j9/KcUUy0oznuKfIiLPqeolJe9pyrIZz72Iz+dj+7ZtXDViGE2aNKVlqzM8jXnz8G74/QW8NG/579af0LQ+d1/bl17/98QRPum9slAzcku8tD57cnkvInMOmd4CBhQuF/O5DBH5XEQ+nzZ1SlixU30+tmzeUrScnZWFz+cL61hlLX60zx0oipdSpw4dO3dh9aqVnsa7qPdf6dGuOZfd9uzv1jdKrcXLD2Vwxe3/5qfMXwH4JXsnjYJqho18tfglqOboljp16rB1azYAW7dmk5ISmWHYov3zj5eaolf3PBsDu4GHgAedaU/Q/GGp6hRVbaWqrcJtNTupeQs2bFhPZuZG8nJzmT9vLu07dgrrWGUtfrTPPWf/fvbt21s0/+knS0hv1syzeF3OOoHRl53LoOufJudAUZsgNatV5o3HruL2R2fz6dfritZv+XU3e/YdoHWLYwAY2qs1b3/oftJu36ETb81+E4C3Zr9Jh46RGYs52j//hAQJeYplour+sIgikkDgecYewBhVXSEi61S1aajHOJAf/niNHy3+kAfuu5eCAj/9+g9kxJUjwz1UmYvvVuxwfi0yN25k9HWjAMj3+zmvR6+w46e0/v0gszMnX8Y5LZtRt1Y1srfv5q6n5jFmWFcqJiexbVeg5+myVeu59p6XGHtFN8Zc3pW1G7YWfb73yMfZumMvp594FFMmXUTlihV4d8m33HD/q4eNH+ogs+PGjOaL5cvZuXMHKXXqcNX/XUPHzp0Ze+MNbN68mQYNG/LAgw9Ts2atUp1/uJeibvz8KyWFV5l7+atNIf/WDD6tUcxmRk+SYtHBRRoDDxN4JWofVQ15UIk/kxTNn+fhr0VIDk2KkVaeR94ONym+suKXkH9rLji1YUgxnFG6Pgc2qWovEWkCvATUAb4ALnYGwK4IPAe0BLYBg1V1fSlPAfD4kSFVzVTV84H/As97GcsYE10e3VO8DlgTtHw/8LCqpgM7gOHO+uHADmf9w85+YYnIc5SqOldVb41ELGNMdLj9nKJzpdkTeMZZFqATUNgNaSbQz5nv6yzjbO8sYTb7R+yRHGNMfEt0/5L/n8DNQHVnuQ6wM6h3XSaBN4vifN0Igd53IrLL2f/X0gaN9R43xpgyojSXz8GP3znT7x43EZFeQLaqfhHJcwCrKRpjXFKaiqKqTgGKexi5LdBHRHoAlYAawCNALRFJcmqLjYFNzv6bgDQg03nbaE0CDS6lZjVFY4wr3HwdgareoqqNVfUY4ELgA1X9G7AQGOTsdikw25mf4yzjbP9Aw3y0xpKiMcYVEXrF6VhgtIisJXDPcJqzfhpQx1k/GhgXbgC7fDbGuEI86sCnqouARc78OqD1YfY5AJzvRjxLisYYV3jQ+hwVlhSNMa6Ik5xoSdEY4w5LisYYE8Sre4qRZknRHFa0/9ffvuzxqMZf8fPOqMY/7ZhaUY0fjhgfESxklhSNMa6Il5G3LSkaY1xhl8/GGBPELp+NMSaI1RSNMSZInNxStKRojHFHnORES4rGGHdYNz9jjAkWHznRkqIxxh3W0BLDlny0mPvvu4cCfwH9B57P8BEZJX8oTuKX53MH+Pdzz/Kf119FRGjW7Dgm3T2ZihUruhojN/cgk8deRX5eLn6/nzPadqL/RRls3fILT94/nr17dnFM+vFk3HgHSRUqALDso/d584WpIMJRTZpx1c13uVomgAnjb2Hxh4tISanDG7Pfdv34JYmTq+f4G2TW7/dz7z138q+nnuE/c+Yyf97b/Lh2bbmIX57PHSArK4sXX3iOWS+/zutvvo2/wM/8/851PU6FCsmMvfcJ7nr8Be587HlWfbGUtd+t4pUZj9O134U88MzrVKlWncXvzgFgy6YNvP3KTG77+1TuffIlhmbc4HqZAPr2G8CTTz/jybFD4dErTiMu7pLi6lUrSUs7msZpaVRITqZ7j54sWrigXMQvz+deyJ/v5+DBA+Tn53Mg5wD16qW6HkNEqFS5ihMvH78/H0FYs/Jzzji7EwBnd+7Jl0s/BODDd2bTudcgqlavAUCNWimulwmgZaszqFGzpifHDkmcZMW4u3zOzsqifoP6RcupPh+rVq4sF/HL87kD+Hw+Lrnscrqf25FKlSrS5qy2nNX2bE9iFfj9TLzuUrI3Z9K55yBSGzSmStXqJCYG/qRq101lx7atQKCmCHD3TSMoKPDTb+gITm51pifliqZ46fsckZqiiJwtIqNFpGsk4pnyafeuXSxauIC57yzg3Q8+Iicnh7lvzS75g2FISEzkrsef56GZb7Huf9+wOXP9Efct8PvJ+mUj4+57kpE3382zj93Lvr17PClXNMVJRdGbpCgiy4LmRwCPE3ih9UQROeILZYLfBTttanFvPzyyVJ+PLZu3FC1nZ2Xh8/nCOlZZi1+ezx1g6dJPaNSoMSkpKVSoUIHOnbuyYsVXnsasWq06J5zckrXfrWL/vj34/YH3tO/4NZvadeoBgVrjaX89h6SkJOrVb4iv0VFk/bLR03JFRZxkRa9qihWC5jOALqo6CegK/O1IH1LVKaraSlVbhdtqeVLzFmzYsJ7MzI3k5eYyf95c2nfsFNaxylr88nzuAA0aNGTlyq/JyclBVfnss09p2vRY1+Ps3rWjqKaXe/AA36xYRsO0JhzfoiXLP/4AgI8XzOW0v7YD4PQ27flu1ZcA7Nm1k6xNG0it38j1ckWblOJfLPPqnmKCiNQmkHRFVbcCqOo+Ecn3KCYASUlJ3HLbBEZmXBG4f9N/IOnpzbwMGTPxy/O5A7Q4+RTO7dKNIRf0JzExieOPP4GB5w92Pc6u7b8y9aE7KSgoQLWA1md35tTWZ9MwrQlPPjCeN/79NEc1PY523foEytWyDd989Rm3XjWYhIRELrj8GqrVcL9BZOxNo/l8+TJ27txBl07tGDnqGgYMdOUFdyGJk1uKSJjviy7+oCLrgQICFWUF2qrqZhGpBnysqqeWdIwD+bhfMFNmePBrWSrleeTtSknhVeW+3rgn5J/aKWnVYzaFelJTVNVjjrCpAOjvRUxjTHTF+mVxqCL6SI6q7gd+imRMY0xkxMvlc9w9p2iMiY44yYnx16PFGBMlLj6SIyJpIrJQRL4VkW9E5DpnfYqIvCciPzhfazvrRUQeFZG1IrJSRE4P9zQsKRpjXOHyIzn5wI2qeiLQBhglIicC44AFqtoMWOAsA5wHNHOmDODJcM/DkqIxxhUJEvpUElXdrKpfOvN7gDVAI6AvMNPZbSbQz5nvCzynAUuBWiLSIKzzCOdDxhjzBx71aBGRY4DTgM8An6pudjZtAQq7TDUCgrsJZTrrSs2SojHGFaW5fA7u0utMh+3C5jzb/DpwvaruDt6mgYesXX+i1VqfjTGuKM0jOao6BSh2gAMRqUAgIb6gqm84q7NEpIHTGaQBkO2s3wSkBX28sbOu1KymaIxxhZtXzyIiwDRgjao+FLRpDnCpM38pMDto/SVOK3QbYFfQZXapWE3RGOMOdx9UbAtcDKwSkRXOuluB+4BXRGQ48DNwgbNtHtADWAvsB4aFG9iSojHGFW4OMquqH3PkNNv5MPsrMMqN2JYUTUyKdpexaA7IUFbFS48WS4rGGHfESVa0pGiMcYWNkmOMMUGifcvDLZYUjTGusKRojDFB7PLZGGOCWE3RGGOCxElOtKRojHGH1RSNMeZ34iMrWlI0xrgilMFjy4K4HCVnyUeL6dOzG726d2Ha1GJHJ4q7+OX53C1+dOOLhD7FNFWNySknTzWcae+BfO3UubP+sG6D7t53UHv16q2r1/wQ1rHKWvzyfO4W37344f7Nbt6Zq6FO0c4vxU1xV1NcvWolaWlH0zgtjQrJyXTv0ZNFCxeUi/jl+dwtfvTje/U6gkjzJCmKyF9FpIYzX1lEJonIWyJyv4jU9CJmoeysLOo3qF+0nOrzkZWV5WXImIlfns/d4kc/fpzkRM9qitMJDPQI8AhQE7jfWTfjSB8Kfm9DNO7HGGPCFy/3FL1qfU5Q1XxnvpWqFr6Y+uOgUXT/IPi9DQfyw3shTarPx5bNW4qWs7Oy8Pl8xXzCXdGMX57P3eJHP77EerYLkVc1xdUiUjgc+Nci0gpARI4D8jyKCcBJzVuwYcN6MjM3kpeby/x5c2nfsZOXIWMmfnk+d4sf/fjxcvnsVU3xCuARERkP/Ap8KiIbCbyX9QqPYgKQlJTELbdNYGTGFRQU+OnXfyDp6c28DBkz8cvzuVv86MePk4oiEni1gUcHDzS2NCGQfDNVNeS7vuFePhtj/pxKSeFV5rbv84f8N5tSNTFmU6inPVqcl1d/7WUMY0xsiJeaonXzM8a4wpKiMcYEsUFmjTEmiNUUjTEmSJzkREuKxhiXxElWtKRojHFFvNxTjLtRcowx0ZEgoU+hEJHuIvK9iKwVkXHelv43lhSNMe5wsZ+fiCQCTwDnAScCQ0TkRC+KfShLisYYV0gp/oWgNbBWVdepai7wEtDX0xNwWFI0xrjC5aHDGhEYK6FQprPOczHb0BJu/8tCIpLhDEUWFRbf4kcrfrRil+ZvVkQygIygVVOi+fMKFs81xYySd7H4Fj8u40f73EukqlNUtVXQdGhC3ASkBS03dtZ5Lp6TojGm7FoONBORJiKSDFwIzIlE4Ji9fDbGlF+qmi8iVwPvAInAdFX9JhKx4zkpRvv+hMW3+OUxtmtUdR4wL9JxPR1k1hhjyhq7p2iMMUHiMilGq3uQE3u6iGSLyOpIxnVip4nIQhH5VkS+EZHrIhy/kogsE5GvnfiTIhk/qByJIvKViLwdhdjrRWSViKwQkc+jEL+WiLwmIt+JyBoROTPSZSjr4u7y2eke9D+gC4EHPpcDQ1T12wjFbwfsBZ5T1eaRiBkUuwHQQFW/FJHqwBdAvwieuwBVVXWviFQAPgauU9WlkYgfVI7RQCughqr2inDs9QRe6/trJOMGxZ8JfKSqzzittlVUdWc0ylJWxWNNMWrdgwBUdTGwPVLxDom9WVW/dOb3AGuIUC8AJ6aq6l5nsYIzRfR/XRFpDPQEnolk3FggIjWBdsA0AFXNtYRYevGYFKPWPSiWiMgxwGnAZxGOmygiK4Bs4D1VjWh84J/AzUBBhOMWUuBdEfnC6bURSU2ArcAM5/bBMyJSNcJlKPPiMSmWeyJSDXgduN55o2LEqKpfVU8l0AOhtYhE7BaCiPQCslX1i0jFPIyzVfV0AqO7jHJup0RKEnA68KSqngbsAyJ6Tz0exGNSjFr3oFjg3Mt7HXhBVd+IVjmcy7aFQPcIhm0L9HHu670EdBKR5yMYH1Xd5HzNBv5D4HZOpGQSeL96Ye38NQJJ0pRCPCbFqHUPijanoWMasEZVH4pC/HoiUsuZr0ygseu7SMVX1VtUtbGqHkPg5/6Bql4UqfgiUtVp4MK5bO0KROwpBFXdAmwUkb84qzoDEWlkiydx16Mlmt2DAETkRaADUFdEMoGJqjotQuHbAhcDq5z7egC3Oj0DIqEBMNN5AiABeEVVI/5YTBT5gP8E/m8iCZilqvMjXIZrgBecCsE6YFiE45d5cfdIjjHG/BnxePlsjDFhs6RojDFBLCkaY0wQS4rGGBPEkqIxxgSxpGiOSET2lrxX0b53iMhNXh3fmEixpGiMMUEsKZpSEZHeIvKZM+DA+yLiC9p8ioh8KiI/iMiIoM+MEZHlIrLycGMsikgDEVnsjEG4WkTOicjJGHMYlhRNaX0MtHEGHHiJwIg0hU4GOgFnAhNEpKGIdAWaEegDfCrQ8jCDJAwF3nEGkjgFWOHlCRhTnLjr5mc81xh42RnQNhn4KWjbbFXNAXJEZCGBRHg2gT7AXzn7VCOQJBcHfW45MN0ZzOJNVV3h7SkYc2RWUzSl9RjwuKq2AK4EKgVtO7TPqAICTFbVU50p/dC+4M7AvO0IjGb0rIhc4l3xjSmeJUVTWjX5bSi2Sw/Z1td5T0sdAoNiLCcwMMflzhiPiEgjEUkN/pCIHA1kqepUAiNm23BXJmrs8tkUp4oz0k+hh4A7gFdFZAfwAYHRngutJDCGYl3gLlX9BfhFRE4APnVGj9kLXERgZO5CHYAxIpLnbLeaookaGyXHGGOC2OWzMcYEsaRojDFBLCkaY0wQS4rGGBPEkqIxxgSxpGiMMUEsKRpjTBBLisYYE+T/Af9p5ZwwhyPvAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 360x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "test_acc, test_f1, test_cm = evaluate(\n",
    "    f'{MODELS_DIR}/{MODEL_NAME}_{MODEL_DTYPE}.tflite', 'tflite', test_data_gen, config)\n",
    "\n",
    "plt.figure(figsize=(5,4));\n",
    "sns.heatmap(\n",
    "    test_cm, annot=True, cmap=\"Blues\", fmt='0.0f', \n",
    "    xticklabels=np.arange(config['dataset'][\"max_people_count\"]+1), \n",
    "    yticklabels=np.arange(config['dataset'][\"max_people_count\"]+1)\n",
    ");\n",
    "plt.xlabel('Labels');\n",
    "plt.ylabel('Predictions');"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7a7fc9aa",
   "metadata": {},
   "source": [
    "## Edge TPU\n",
    "\n",
    "- [Edge TPU Compiler](https://coral.ai/docs/edgetpu/compiler/)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "79fd91cd",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Edge TPU Compiler version 16.0.384591198\r\n",
      "Started a compilation timeout timer of 180 seconds.\r\n",
      "\r\n",
      "Model compiled successfully in 1 ms.\r\n",
      "\r\n",
      "Input model: ./data/model/model_fp32.tflite\r\n",
      "Input size: 187.26KiB\r\n",
      "Output model: model_fp32_edgetpu.tflite\r\n",
      "Output size: 186.26KiB\r\n",
      "On-chip memory used for caching model parameters: 0.00B\r\n",
      "On-chip memory remaining for caching model parameters: 0.00B\r\n",
      "Off-chip memory used for streaming uncached model parameters: 0.00B\r\n",
      "Number of Edge TPU subgraphs: 0\r\n",
      "Total number of operations: 13\r\n",
      "Operation log: model_fp32_edgetpu.log\r\n",
      "\r\n",
      "Model successfully compiled but not all operations are supported by the Edge TPU. A percentage of the model will instead run on the CPU, which is slower. If possible, consider updating your model to use only operations supported by the Edge TPU. For details, visit g.co/coral/model-reqs.\r\n",
      "Number of operations that will run on Edge TPU: 0\r\n",
      "Number of operations that will run on CPU: 13\r\n",
      "\r\n",
      "Operator                       Count      Status\r\n",
      "\r\n",
      "RESIZE_NEAREST_NEIGHBOR        1          Operation is working on an unsupported data type\r\n",
      "MAX_POOL_2D                    1          Operation is working on an unsupported data type\r\n",
      "CONCATENATION                  1          Operation is working on an unsupported data type\r\n",
      "CONV_2D                        10         Operation is working on an unsupported data type\r\n",
      "Compilation child process completed within timeout period.\r\n",
      "Compilation succeeded! \r\n"
     ]
    }
   ],
   "source": [
    "!edgetpu_compiler -s {MODELS_DIR}/{MODEL_NAME}_{MODEL_DTYPE}.tflite -o {MODELS_DIR}"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1301c750",
   "metadata": {},
   "source": [
    "## TF Lite Micro conversion\n",
    "\n",
    "- [TF Lite Micro build and convert models](https://www.tensorflow.org/lite/microcontrollers/build_convert)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "c0f367f2",
   "metadata": {},
   "outputs": [],
   "source": [
    "!xxd -i {MODELS_DIR}/{MODEL_NAME}_{MODEL_DTYPE}.tflite > {MODELS_DIR}/{MODEL_NAME}_{MODEL_DTYPE}.cc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "c59be4aa",
   "metadata": {},
   "outputs": [],
   "source": [
    "!sed -i 's/'{MODELS_DIR.replace('.', '_').replace('/','_')}'_'{MODEL_NAME}'_'{MODEL_DTYPE}'_tflite/g_model/g' {MODELS_DIR}/{MODEL_NAME}_{MODEL_DTYPE}.cc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "aa5f1430",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  0x34, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00,\r\n",
      "  0xcc, 0xff, 0xff, 0xff, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02,\r\n",
      "  0x0c, 0x00, 0x10, 0x00, 0x0f, 0x00, 0x00, 0x00, 0x08, 0x00, 0x04, 0x00,\r\n",
      "  0x0c, 0x00, 0x00, 0x00, 0x61, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00,\r\n",
      "  0x00, 0x00, 0x00, 0x61, 0xf4, 0xff, 0xff, 0xff, 0x11, 0x00, 0x00, 0x00,\r\n",
      "  0x00, 0x00, 0x00, 0x11, 0x0c, 0x00, 0x0c, 0x00, 0x0b, 0x00, 0x00, 0x00,\r\n",
      "  0x00, 0x00, 0x04, 0x00, 0x0c, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00,\r\n",
      "  0x00, 0x00, 0x00, 0x03\r\n",
      "};\r\n",
      "unsigned int g_model_len = 191752;\r\n"
     ]
    }
   ],
   "source": [
    "!cat {MODELS_DIR}/{MODEL_NAME}_{MODEL_DTYPE}.cc | tail"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "205cbbe8",
   "metadata": {},
   "source": [
    "## Convert to ONNX\n",
    "\n",
    "- [tf2onnx](https://github.com/onnx/tensorflow-onnx) package to convert from tensorflow to ONNX format"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "fad2f20d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "/home/mateusz/miniconda3/envs/thermo/lib/python3.9/runpy.py:127: RuntimeWarning: 'tf2onnx.convert' found in sys.modules after import of package 'tf2onnx', but prior to execution of 'tf2onnx.convert'; this may result in unpredictable behaviour\n",
      "  warn(RuntimeWarning(msg))\n",
      "2022-12-08 15:56:58,559 - INFO - Using tensorflow=2.7.0, onnx=1.12.0, tf2onnx=1.13.0/2c1db5\n",
      "2022-12-08 15:56:58,559 - INFO - Using opset <onnx, 13>\n",
      "2022-12-08 15:56:58,606 - INFO - Optimizing ONNX model\n",
      "2022-12-08 15:56:58,711 - INFO - After optimization: Cast -1 (1->0), Const +1 (25->26), Identity -1 (1->0), Reshape +2 (0->2), Transpose -44 (44->0)\n",
      "2022-12-08 15:56:58,714 - INFO - \n",
      "2022-12-08 15:56:58,714 - INFO - Successfully converted TensorFlow model ./data/model/model_fp32.tflite to ONNX\n",
      "2022-12-08 15:56:58,714 - INFO - Model inputs: ['serving_default_input_1:0']\n",
      "2022-12-08 15:56:58,714 - INFO - Model outputs: ['StatefulPartitionedCall:0']\n",
      "2022-12-08 15:56:58,714 - INFO - ONNX model is saved at ./data/model/model.onnx\n"
     ]
    }
   ],
   "source": [
    "!python -m tf2onnx.convert --tflite {MODELS_DIR}/{MODEL_NAME}_fp32.tflite --output {MODELS_DIR}/{MODEL_NAME}.onnx"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "91939e99",
   "metadata": {},
   "source": [
    "## Convert to OpenVINO\n",
    "\n",
    "- [OpenVINO toolkit](https://www.intel.com/content/www/us/en/developer/tools/openvino-toolkit/overview.html) to optimize and convert ONNX model to OpenVINO format"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "7f91b772",
   "metadata": {},
   "outputs": [],
   "source": [
    "OpenVINO_PATH = '~/intel/openvino_2021'\n",
    "DATA_TYPE = 'FP32' # FP32 FP16"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "81ecc7df",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[setupvars.sh] OpenVINO environment initialized\r\n"
     ]
    }
   ],
   "source": [
    "!source {OpenVINO_PATH}/bin/setupvars.sh"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "b04e6eed",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model Optimizer arguments:\n",
      "Common parameters:\n",
      "\t- Path to the Input Model: \t/home/mateusz/Research/thermo-hardware-benchmark/thermo/./data/model/model.onnx\n",
      "\t- Path for generated IR: \t/home/mateusz/Research/thermo-hardware-benchmark/thermo/./data/model\n",
      "\t- IR output name: \tmodel_FP32\n",
      "\t- Log level: \tERROR\n",
      "\t- Batch: \t1\n",
      "\t- Input layers: \tNot specified, inherited from the model\n",
      "\t- Output layers: \tNot specified, inherited from the model\n",
      "\t- Input shapes: \tNot specified, inherited from the model\n",
      "\t- Mean values: \tNot specified\n",
      "\t- Scale values: \tNot specified\n",
      "\t- Scale factor: \tNot specified\n",
      "\t- Precision of IR: \tFP32\n",
      "\t- Enable fusing: \tTrue\n",
      "\t- Enable grouped convolutions fusing: \tTrue\n",
      "\t- Move mean values to preprocess section: \tNone\n",
      "\t- Reverse input channels: \tFalse\n",
      "ONNX specific parameters:\n",
      "\t- Inference Engine found in: \t/home/mateusz/intel/openvino_2021.4.582/python/python3.9/openvino\n",
      "Inference Engine version: \t2021.4.0-3839-cd81789d294-releases/2021/4\n",
      "Model Optimizer version: \t2021.4.0-3839-cd81789d294-releases/2021/4\n",
      "/home/mateusz/intel/openvino_2021.4.582/deployment_tools/model_optimizer/mo/utils/versions_checker.py:170: DeprecationWarning: distutils Version classes are deprecated. Use packaging.version instead.\n",
      "  req_ver = LooseVersion(required_v)\n",
      "/home/mateusz/miniconda3/envs/thermo/lib/python3.9/site-packages/setuptools/_distutils/version.py:346: DeprecationWarning: distutils Version classes are deprecated. Use packaging.version instead.\n",
      "  other = LooseVersion(other)\n",
      "/home/mateusz/intel/openvino_2021.4.582/deployment_tools/model_optimizer/mo/utils/versions_checker.py:170: DeprecationWarning: distutils Version classes are deprecated. Use packaging.version instead.\n",
      "  req_ver = LooseVersion(required_v)\n",
      "/home/mateusz/miniconda3/envs/thermo/lib/python3.9/site-packages/setuptools/_distutils/version.py:346: DeprecationWarning: distutils Version classes are deprecated. Use packaging.version instead.\n",
      "  other = LooseVersion(other)\n",
      "[ WARNING ]  \n",
      "Detected not satisfied dependencies:\n",
      "\tnumpy: installed: 1.20.3, required: < 1.20\n",
      "\n",
      "Please install required versions of components or use install_prerequisites script\n",
      "/home/mateusz/intel/openvino_2021.4.582/deployment_tools/model_optimizer/install_prerequisites/install_prerequisites_onnx.sh\n",
      "Note that install_prerequisites scripts may install additional components.\n",
      "/home/mateusz/intel/openvino_2021.4.582/deployment_tools/model_optimizer/extensions/analysis/boolean_input.py:13: DeprecationWarning: `np.bool` is a deprecated alias for the builtin `bool`. To silence this warning, use `bool` by itself. Doing this will not modify any behavior and is safe. If you specifically wanted the numpy scalar type, use `np.bool_` here.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  nodes = graph.get_op_nodes(op='Parameter', data_type=np.bool)\n",
      "/home/mateusz/intel/openvino_2021.4.582/deployment_tools/model_optimizer/mo/front/common/partial_infer/concat.py:36: DeprecationWarning: `np.bool` is a deprecated alias for the builtin `bool`. To silence this warning, use `bool` by itself. Doing this will not modify any behavior and is safe. If you specifically wanted the numpy scalar type, use `np.bool_` here.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  mask = np.zeros_like(shape, dtype=np.bool)\n",
      "[ SUCCESS ] Generated IR version 10 model.\n",
      "[ SUCCESS ] XML file: /home/mateusz/Research/thermo-hardware-benchmark/thermo/data/model/model_FP32.xml\n",
      "[ SUCCESS ] BIN file: /home/mateusz/Research/thermo-hardware-benchmark/thermo/data/model/model_FP32.bin\n",
      "[ SUCCESS ] Total execution time: 2.82 seconds. \n",
      "[ SUCCESS ] Memory consumed: 85 MB. \n",
      "It's been a while, check for a new version of Intel(R) Distribution of OpenVINO(TM) toolkit here https://software.intel.com/content/www/us/en/develop/tools/openvino-toolkit/download.html?cid=other&source=prod&campid=ww_2021_bu_IOTG_OpenVINO-2021-4-LTS&content=upg_all&medium=organic or on the GitHub*\n"
     ]
    }
   ],
   "source": [
    "!python {OpenVINO_PATH}/deployment_tools/model_optimizer/mo.py \\\n",
    "    --input_model {MODELS_DIR}/{MODEL_NAME}.onnx \\\n",
    "    --model_name {MODEL_NAME}_{DATA_TYPE} \\\n",
    "    --output_dir {MODELS_DIR} \\\n",
    "    --data_type {DATA_TYPE} \\\n",
    "    --batch 1"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}