ClassificationTFLuNano.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "mount_file_id": "16H6NFJle0YxLAwYlwKbI7ZPpa_fycGM7",
      "authorship_tag": "ABX9TyP83jmqe+KGBlvTzQ8c0fFk",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/AdamClarkStandke/TinyMachineLearning/blob/main/ClassificationTFLuNano.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "Sf0cvKQp5n-i"
      },
      "outputs": [],
      "source": [
        "%tensorflow_version 2.x"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "import pathlib\n",
        "import tensorflow as tf\n",
        "import tensorflow_datasets as tfds\n",
        "import zipfile\n",
        "from tensorflow.keras.applications.mobilenet_v2 import MobileNetV2\n",
        "from tensorflow.keras.models import Model"
      ],
      "metadata": {
        "id": "QFglS7gN6LjK"
      },
      "execution_count": 1,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "MODEL_ALPHA = 0.35\n",
        "MODEL_INPUT_WIDTH = 48\n",
        "MODEL_INPUT_HEIGHT = 48\n",
        "TFL_MODEL_FILE = \"recognize.tflite\"\n",
        "TFL_MODEL_HEADER_FILE = \"recognize_model.h\"\n",
        "TF_MODEL = \"drive/MyDrive/recognize/\""
      ],
      "metadata": {
        "id": "I76kxDk86Qt0"
      },
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "with zipfile.ZipFile(\"dataset.zip\", 'r') as zip_ref:\n",
        "    zip_ref.extractall(\".\")\n",
        "data_dir = \"dataset\""
      ],
      "metadata": {
        "id": "-C6aHNuW6vNr"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "train_ds = tf.keras.utils.image_dataset_from_directory(\n",
        "  data_dir,\n",
        "  validation_split=0.2,\n",
        "  subset=\"training\",\n",
        "  seed=123,\n",
        "  interpolation=\"bilinear\",\n",
        "  image_size=(MODEL_INPUT_WIDTH, MODEL_INPUT_HEIGHT)\n",
        "  )\n",
        "\n",
        "\n",
        "val_ds = tf.keras.utils.image_dataset_from_directory(\n",
        "  data_dir,\n",
        "  validation_split=0.2,\n",
        "  subset=\"validation\",\n",
        "  seed=123,\n",
        "  interpolation=\"bilinear\",\n",
        "  image_size=(MODEL_INPUT_WIDTH, MODEL_INPUT_HEIGHT)\n",
        "  )"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "CWpvPB9e61o3",
        "outputId": "56819bd1-37f5-4ffb-80c0-405c0833352f"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Found 40 files belonging to 2 classes.\n",
            "Using 32 files for training.\n",
            "Found 40 files belonging to 2 classes.\n",
            "Using 8 files for validation.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "class_names = train_ds.class_names\n",
        "num_classes = len(class_names)\n",
        "print(class_names)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "jqHQ6Xvp8HIf",
        "outputId": "5cb2062d-4643-45c2-bf07-da484dcae2e3"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "['bathroom', 'kitchen']\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "rescale = tf.keras.layers.Rescaling(1./255, offset= -1)\n",
        "train_ds = train_ds.map(lambda x, y: (rescale(x), y))\n",
        "val_ds   = val_ds.map(lambda x, y: (rescale(x), y))"
      ],
      "metadata": {
        "id": "b3yOji1I8LYf"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "base_model = MobileNetV2(input_shape=(MODEL_INPUT_WIDTH, MODEL_INPUT_HEIGHT, 3),\n",
        "                         include_top=False,\n",
        "                         weights='imagenet',\n",
        "                         alpha=0.35)\n",
        "base_model.trainable = False\n",
        "\n",
        "feat_extr = base_model"
      ],
      "metadata": {
        "id": "O1uXdRdu8Vss"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "augmen = tf.keras.Sequential([\n",
        "  tf.keras.layers.experimental.preprocessing.RandomRotation(0.2),\n",
        "  tf.keras.layers.experimental.preprocessing.RandomFlip('horizontal'),\n",
        "])\n",
        "\n",
        "train_ds = train_ds.map(lambda x, y: (augmen(x), y))\n",
        "val_ds   = val_ds.map(lambda x, y: (augmen(x), y))"
      ],
      "metadata": {
        "id": "-LW3N22F8mM_"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "global_avg_layer = tf.keras.layers.GlobalAveragePooling2D()\n",
        "dense_layer = tf.keras.layers.Dense(num_classes, activation='softmax')"
      ],
      "metadata": {
        "id": "1AxzEJGX80uY"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "inputs = tf.keras.Input(shape=(MODEL_INPUT_WIDTH, MODEL_INPUT_HEIGHT, 3))\n",
        "x = global_avg_layer(feat_extr.layers[-1].output)\n",
        "x = tf.keras.layers.Dropout(0.2)(x)\n",
        "outputs = dense_layer(x)\n",
        "model = tf.keras.Model(inputs=feat_extr.inputs, outputs=outputs)"
      ],
      "metadata": {
        "id": "fFDJK6ko9AwG"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "lr = 0.0005\n",
        "model.compile(\n",
        "  optimizer=tf.keras.optimizers.Adam(learning_rate=lr),\n",
        "  loss=tf.losses.SparseCategoricalCrossentropy(from_logits=False),\n",
        "  metrics=['accuracy'])\n",
        "     "
      ],
      "metadata": {
        "id": "PmK8Mn5-9Dce"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model.summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "MlRSx4Wf9Le6",
        "outputId": "ff6a5328-8363-4d8a-abb3-c79720e2de92"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"model\"\n",
            "__________________________________________________________________________________________________\n",
            " Layer (type)                   Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            " input_1 (InputLayer)           [(None, 48, 48, 3)]  0           []                               \n",
            "                                                                                                  \n",
            " Conv1 (Conv2D)                 (None, 24, 24, 16)   432         ['input_1[0][0]']                \n",
            "                                                                                                  \n",
            " bn_Conv1 (BatchNormalization)  (None, 24, 24, 16)   64          ['Conv1[0][0]']                  \n",
            "                                                                                                  \n",
            " Conv1_relu (ReLU)              (None, 24, 24, 16)   0           ['bn_Conv1[0][0]']               \n",
            "                                                                                                  \n",
            " expanded_conv_depthwise (Depth  (None, 24, 24, 16)  144         ['Conv1_relu[0][0]']             \n",
            " wiseConv2D)                                                                                      \n",
            "                                                                                                  \n",
            " expanded_conv_depthwise_BN (Ba  (None, 24, 24, 16)  64          ['expanded_conv_depthwise[0][0]']\n",
            " tchNormalization)                                                                                \n",
            "                                                                                                  \n",
            " expanded_conv_depthwise_relu (  (None, 24, 24, 16)  0           ['expanded_conv_depthwise_BN[0][0\n",
            " ReLU)                                                           ]']                              \n",
            "                                                                                                  \n",
            " expanded_conv_project (Conv2D)  (None, 24, 24, 8)   128         ['expanded_conv_depthwise_relu[0]\n",
            "                                                                 [0]']                            \n",
            "                                                                                                  \n",
            " expanded_conv_project_BN (Batc  (None, 24, 24, 8)   32          ['expanded_conv_project[0][0]']  \n",
            " hNormalization)                                                                                  \n",
            "                                                                                                  \n",
            " block_1_expand (Conv2D)        (None, 24, 24, 48)   384         ['expanded_conv_project_BN[0][0]'\n",
            "                                                                 ]                                \n",
            "                                                                                                  \n",
            " block_1_expand_BN (BatchNormal  (None, 24, 24, 48)  192         ['block_1_expand[0][0]']         \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " block_1_expand_relu (ReLU)     (None, 24, 24, 48)   0           ['block_1_expand_BN[0][0]']      \n",
            "                                                                                                  \n",
            " block_1_pad (ZeroPadding2D)    (None, 25, 25, 48)   0           ['block_1_expand_relu[0][0]']    \n",
            "                                                                                                  \n",
            " block_1_depthwise (DepthwiseCo  (None, 12, 12, 48)  432         ['block_1_pad[0][0]']            \n",
            " nv2D)                                                                                            \n",
            "                                                                                                  \n",
            " block_1_depthwise_BN (BatchNor  (None, 12, 12, 48)  192         ['block_1_depthwise[0][0]']      \n",
            " malization)                                                                                      \n",
            "                                                                                                  \n",
            " block_1_depthwise_relu (ReLU)  (None, 12, 12, 48)   0           ['block_1_depthwise_BN[0][0]']   \n",
            "                                                                                                  \n",
            " block_1_project (Conv2D)       (None, 12, 12, 8)    384         ['block_1_depthwise_relu[0][0]'] \n",
            "                                                                                                  \n",
            " block_1_project_BN (BatchNorma  (None, 12, 12, 8)   32          ['block_1_project[0][0]']        \n",
            " lization)                                                                                        \n",
            "                                                                                                  \n",
            " block_2_expand (Conv2D)        (None, 12, 12, 48)   384         ['block_1_project_BN[0][0]']     \n",
            "                                                                                                  \n",
            " block_2_expand_BN (BatchNormal  (None, 12, 12, 48)  192         ['block_2_expand[0][0]']         \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " block_2_expand_relu (ReLU)     (None, 12, 12, 48)   0           ['block_2_expand_BN[0][0]']      \n",
            "                                                                                                  \n",
            " block_2_depthwise (DepthwiseCo  (None, 12, 12, 48)  432         ['block_2_expand_relu[0][0]']    \n",
            " nv2D)                                                                                            \n",
            "                                                                                                  \n",
            " block_2_depthwise_BN (BatchNor  (None, 12, 12, 48)  192         ['block_2_depthwise[0][0]']      \n",
            " malization)                                                                                      \n",
            "                                                                                                  \n",
            " block_2_depthwise_relu (ReLU)  (None, 12, 12, 48)   0           ['block_2_depthwise_BN[0][0]']   \n",
            "                                                                                                  \n",
            " block_2_project (Conv2D)       (None, 12, 12, 8)    384         ['block_2_depthwise_relu[0][0]'] \n",
            "                                                                                                  \n",
            " block_2_project_BN (BatchNorma  (None, 12, 12, 8)   32          ['block_2_project[0][0]']        \n",
            " lization)                                                                                        \n",
            "                                                                                                  \n",
            " block_2_add (Add)              (None, 12, 12, 8)    0           ['block_1_project_BN[0][0]',     \n",
            "                                                                  'block_2_project_BN[0][0]']     \n",
            "                                                                                                  \n",
            " block_3_expand (Conv2D)        (None, 12, 12, 48)   384         ['block_2_add[0][0]']            \n",
            "                                                                                                  \n",
            " block_3_expand_BN (BatchNormal  (None, 12, 12, 48)  192         ['block_3_expand[0][0]']         \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " block_3_expand_relu (ReLU)     (None, 12, 12, 48)   0           ['block_3_expand_BN[0][0]']      \n",
            "                                                                                                  \n",
            " block_3_pad (ZeroPadding2D)    (None, 13, 13, 48)   0           ['block_3_expand_relu[0][0]']    \n",
            "                                                                                                  \n",
            " block_3_depthwise (DepthwiseCo  (None, 6, 6, 48)    432         ['block_3_pad[0][0]']            \n",
            " nv2D)                                                                                            \n",
            "                                                                                                  \n",
            " block_3_depthwise_BN (BatchNor  (None, 6, 6, 48)    192         ['block_3_depthwise[0][0]']      \n",
            " malization)                                                                                      \n",
            "                                                                                                  \n",
            " block_3_depthwise_relu (ReLU)  (None, 6, 6, 48)     0           ['block_3_depthwise_BN[0][0]']   \n",
            "                                                                                                  \n",
            " block_3_project (Conv2D)       (None, 6, 6, 16)     768         ['block_3_depthwise_relu[0][0]'] \n",
            "                                                                                                  \n",
            " block_3_project_BN (BatchNorma  (None, 6, 6, 16)    64          ['block_3_project[0][0]']        \n",
            " lization)                                                                                        \n",
            "                                                                                                  \n",
            " block_4_expand (Conv2D)        (None, 6, 6, 96)     1536        ['block_3_project_BN[0][0]']     \n",
            "                                                                                                  \n",
            " block_4_expand_BN (BatchNormal  (None, 6, 6, 96)    384         ['block_4_expand[0][0]']         \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " block_4_expand_relu (ReLU)     (None, 6, 6, 96)     0           ['block_4_expand_BN[0][0]']      \n",
            "                                                                                                  \n",
            " block_4_depthwise (DepthwiseCo  (None, 6, 6, 96)    864         ['block_4_expand_relu[0][0]']    \n",
            " nv2D)                                                                                            \n",
            "                                                                                                  \n",
            " block_4_depthwise_BN (BatchNor  (None, 6, 6, 96)    384         ['block_4_depthwise[0][0]']      \n",
            " malization)                                                                                      \n",
            "                                                                                                  \n",
            " block_4_depthwise_relu (ReLU)  (None, 6, 6, 96)     0           ['block_4_depthwise_BN[0][0]']   \n",
            "                                                                                                  \n",
            " block_4_project (Conv2D)       (None, 6, 6, 16)     1536        ['block_4_depthwise_relu[0][0]'] \n",
            "                                                                                                  \n",
            " block_4_project_BN (BatchNorma  (None, 6, 6, 16)    64          ['block_4_project[0][0]']        \n",
            " lization)                                                                                        \n",
            "                                                                                                  \n",
            " block_4_add (Add)              (None, 6, 6, 16)     0           ['block_3_project_BN[0][0]',     \n",
            "                                                                  'block_4_project_BN[0][0]']     \n",
            "                                                                                                  \n",
            " block_5_expand (Conv2D)        (None, 6, 6, 96)     1536        ['block_4_add[0][0]']            \n",
            "                                                                                                  \n",
            " block_5_expand_BN (BatchNormal  (None, 6, 6, 96)    384         ['block_5_expand[0][0]']         \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " block_5_expand_relu (ReLU)     (None, 6, 6, 96)     0           ['block_5_expand_BN[0][0]']      \n",
            "                                                                                                  \n",
            " block_5_depthwise (DepthwiseCo  (None, 6, 6, 96)    864         ['block_5_expand_relu[0][0]']    \n",
            " nv2D)                                                                                            \n",
            "                                                                                                  \n",
            " block_5_depthwise_BN (BatchNor  (None, 6, 6, 96)    384         ['block_5_depthwise[0][0]']      \n",
            " malization)                                                                                      \n",
            "                                                                                                  \n",
            " block_5_depthwise_relu (ReLU)  (None, 6, 6, 96)     0           ['block_5_depthwise_BN[0][0]']   \n",
            "                                                                                                  \n",
            " block_5_project (Conv2D)       (None, 6, 6, 16)     1536        ['block_5_depthwise_relu[0][0]'] \n",
            "                                                                                                  \n",
            " block_5_project_BN (BatchNorma  (None, 6, 6, 16)    64          ['block_5_project[0][0]']        \n",
            " lization)                                                                                        \n",
            "                                                                                                  \n",
            " block_5_add (Add)              (None, 6, 6, 16)     0           ['block_4_add[0][0]',            \n",
            "                                                                  'block_5_project_BN[0][0]']     \n",
            "                                                                                                  \n",
            " block_6_expand (Conv2D)        (None, 6, 6, 96)     1536        ['block_5_add[0][0]']            \n",
            "                                                                                                  \n",
            " block_6_expand_BN (BatchNormal  (None, 6, 6, 96)    384         ['block_6_expand[0][0]']         \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " block_6_expand_relu (ReLU)     (None, 6, 6, 96)     0           ['block_6_expand_BN[0][0]']      \n",
            "                                                                                                  \n",
            " block_6_pad (ZeroPadding2D)    (None, 7, 7, 96)     0           ['block_6_expand_relu[0][0]']    \n",
            "                                                                                                  \n",
            " block_6_depthwise (DepthwiseCo  (None, 3, 3, 96)    864         ['block_6_pad[0][0]']            \n",
            " nv2D)                                                                                            \n",
            "                                                                                                  \n",
            " block_6_depthwise_BN (BatchNor  (None, 3, 3, 96)    384         ['block_6_depthwise[0][0]']      \n",
            " malization)                                                                                      \n",
            "                                                                                                  \n",
            " block_6_depthwise_relu (ReLU)  (None, 3, 3, 96)     0           ['block_6_depthwise_BN[0][0]']   \n",
            "                                                                                                  \n",
            " block_6_project (Conv2D)       (None, 3, 3, 24)     2304        ['block_6_depthwise_relu[0][0]'] \n",
            "                                                                                                  \n",
            " block_6_project_BN (BatchNorma  (None, 3, 3, 24)    96          ['block_6_project[0][0]']        \n",
            " lization)                                                                                        \n",
            "                                                                                                  \n",
            " block_7_expand (Conv2D)        (None, 3, 3, 144)    3456        ['block_6_project_BN[0][0]']     \n",
            "                                                                                                  \n",
            " block_7_expand_BN (BatchNormal  (None, 3, 3, 144)   576         ['block_7_expand[0][0]']         \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " block_7_expand_relu (ReLU)     (None, 3, 3, 144)    0           ['block_7_expand_BN[0][0]']      \n",
            "                                                                                                  \n",
            " block_7_depthwise (DepthwiseCo  (None, 3, 3, 144)   1296        ['block_7_expand_relu[0][0]']    \n",
            " nv2D)                                                                                            \n",
            "                                                                                                  \n",
            " block_7_depthwise_BN (BatchNor  (None, 3, 3, 144)   576         ['block_7_depthwise[0][0]']      \n",
            " malization)                                                                                      \n",
            "                                                                                                  \n",
            " block_7_depthwise_relu (ReLU)  (None, 3, 3, 144)    0           ['block_7_depthwise_BN[0][0]']   \n",
            "                                                                                                  \n",
            " block_7_project (Conv2D)       (None, 3, 3, 24)     3456        ['block_7_depthwise_relu[0][0]'] \n",
            "                                                                                                  \n",
            " block_7_project_BN (BatchNorma  (None, 3, 3, 24)    96          ['block_7_project[0][0]']        \n",
            " lization)                                                                                        \n",
            "                                                                                                  \n",
            " block_7_add (Add)              (None, 3, 3, 24)     0           ['block_6_project_BN[0][0]',     \n",
            "                                                                  'block_7_project_BN[0][0]']     \n",
            "                                                                                                  \n",
            " block_8_expand (Conv2D)        (None, 3, 3, 144)    3456        ['block_7_add[0][0]']            \n",
            "                                                                                                  \n",
            " block_8_expand_BN (BatchNormal  (None, 3, 3, 144)   576         ['block_8_expand[0][0]']         \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " block_8_expand_relu (ReLU)     (None, 3, 3, 144)    0           ['block_8_expand_BN[0][0]']      \n",
            "                                                                                                  \n",
            " block_8_depthwise (DepthwiseCo  (None, 3, 3, 144)   1296        ['block_8_expand_relu[0][0]']    \n",
            " nv2D)                                                                                            \n",
            "                                                                                                  \n",
            " block_8_depthwise_BN (BatchNor  (None, 3, 3, 144)   576         ['block_8_depthwise[0][0]']      \n",
            " malization)                                                                                      \n",
            "                                                                                                  \n",
            " block_8_depthwise_relu (ReLU)  (None, 3, 3, 144)    0           ['block_8_depthwise_BN[0][0]']   \n",
            "                                                                                                  \n",
            " block_8_project (Conv2D)       (None, 3, 3, 24)     3456        ['block_8_depthwise_relu[0][0]'] \n",
            "                                                                                                  \n",
            " block_8_project_BN (BatchNorma  (None, 3, 3, 24)    96          ['block_8_project[0][0]']        \n",
            " lization)                                                                                        \n",
            "                                                                                                  \n",
            " block_8_add (Add)              (None, 3, 3, 24)     0           ['block_7_add[0][0]',            \n",
            "                                                                  'block_8_project_BN[0][0]']     \n",
            "                                                                                                  \n",
            " block_9_expand (Conv2D)        (None, 3, 3, 144)    3456        ['block_8_add[0][0]']            \n",
            "                                                                                                  \n",
            " block_9_expand_BN (BatchNormal  (None, 3, 3, 144)   576         ['block_9_expand[0][0]']         \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " block_9_expand_relu (ReLU)     (None, 3, 3, 144)    0           ['block_9_expand_BN[0][0]']      \n",
            "                                                                                                  \n",
            " block_9_depthwise (DepthwiseCo  (None, 3, 3, 144)   1296        ['block_9_expand_relu[0][0]']    \n",
            " nv2D)                                                                                            \n",
            "                                                                                                  \n",
            " block_9_depthwise_BN (BatchNor  (None, 3, 3, 144)   576         ['block_9_depthwise[0][0]']      \n",
            " malization)                                                                                      \n",
            "                                                                                                  \n",
            " block_9_depthwise_relu (ReLU)  (None, 3, 3, 144)    0           ['block_9_depthwise_BN[0][0]']   \n",
            "                                                                                                  \n",
            " block_9_project (Conv2D)       (None, 3, 3, 24)     3456        ['block_9_depthwise_relu[0][0]'] \n",
            "                                                                                                  \n",
            " block_9_project_BN (BatchNorma  (None, 3, 3, 24)    96          ['block_9_project[0][0]']        \n",
            " lization)                                                                                        \n",
            "                                                                                                  \n",
            " block_9_add (Add)              (None, 3, 3, 24)     0           ['block_8_add[0][0]',            \n",
            "                                                                  'block_9_project_BN[0][0]']     \n",
            "                                                                                                  \n",
            " block_10_expand (Conv2D)       (None, 3, 3, 144)    3456        ['block_9_add[0][0]']            \n",
            "                                                                                                  \n",
            " block_10_expand_BN (BatchNorma  (None, 3, 3, 144)   576         ['block_10_expand[0][0]']        \n",
            " lization)                                                                                        \n",
            "                                                                                                  \n",
            " block_10_expand_relu (ReLU)    (None, 3, 3, 144)    0           ['block_10_expand_BN[0][0]']     \n",
            "                                                                                                  \n",
            " block_10_depthwise (DepthwiseC  (None, 3, 3, 144)   1296        ['block_10_expand_relu[0][0]']   \n",
            " onv2D)                                                                                           \n",
            "                                                                                                  \n",
            " block_10_depthwise_BN (BatchNo  (None, 3, 3, 144)   576         ['block_10_depthwise[0][0]']     \n",
            " rmalization)                                                                                     \n",
            "                                                                                                  \n",
            " block_10_depthwise_relu (ReLU)  (None, 3, 3, 144)   0           ['block_10_depthwise_BN[0][0]']  \n",
            "                                                                                                  \n",
            " block_10_project (Conv2D)      (None, 3, 3, 32)     4608        ['block_10_depthwise_relu[0][0]']\n",
            "                                                                                                  \n",
            " block_10_project_BN (BatchNorm  (None, 3, 3, 32)    128         ['block_10_project[0][0]']       \n",
            " alization)                                                                                       \n",
            "                                                                                                  \n",
            " block_11_expand (Conv2D)       (None, 3, 3, 192)    6144        ['block_10_project_BN[0][0]']    \n",
            "                                                                                                  \n",
            " block_11_expand_BN (BatchNorma  (None, 3, 3, 192)   768         ['block_11_expand[0][0]']        \n",
            " lization)                                                                                        \n",
            "                                                                                                  \n",
            " block_11_expand_relu (ReLU)    (None, 3, 3, 192)    0           ['block_11_expand_BN[0][0]']     \n",
            "                                                                                                  \n",
            " block_11_depthwise (DepthwiseC  (None, 3, 3, 192)   1728        ['block_11_expand_relu[0][0]']   \n",
            " onv2D)                                                                                           \n",
            "                                                                                                  \n",
            " block_11_depthwise_BN (BatchNo  (None, 3, 3, 192)   768         ['block_11_depthwise[0][0]']     \n",
            " rmalization)                                                                                     \n",
            "                                                                                                  \n",
            " block_11_depthwise_relu (ReLU)  (None, 3, 3, 192)   0           ['block_11_depthwise_BN[0][0]']  \n",
            "                                                                                                  \n",
            " block_11_project (Conv2D)      (None, 3, 3, 32)     6144        ['block_11_depthwise_relu[0][0]']\n",
            "                                                                                                  \n",
            " block_11_project_BN (BatchNorm  (None, 3, 3, 32)    128         ['block_11_project[0][0]']       \n",
            " alization)                                                                                       \n",
            "                                                                                                  \n",
            " block_11_add (Add)             (None, 3, 3, 32)     0           ['block_10_project_BN[0][0]',    \n",
            "                                                                  'block_11_project_BN[0][0]']    \n",
            "                                                                                                  \n",
            " block_12_expand (Conv2D)       (None, 3, 3, 192)    6144        ['block_11_add[0][0]']           \n",
            "                                                                                                  \n",
            " block_12_expand_BN (BatchNorma  (None, 3, 3, 192)   768         ['block_12_expand[0][0]']        \n",
            " lization)                                                                                        \n",
            "                                                                                                  \n",
            " block_12_expand_relu (ReLU)    (None, 3, 3, 192)    0           ['block_12_expand_BN[0][0]']     \n",
            "                                                                                                  \n",
            " block_12_depthwise (DepthwiseC  (None, 3, 3, 192)   1728        ['block_12_expand_relu[0][0]']   \n",
            " onv2D)                                                                                           \n",
            "                                                                                                  \n",
            " block_12_depthwise_BN (BatchNo  (None, 3, 3, 192)   768         ['block_12_depthwise[0][0]']     \n",
            " rmalization)                                                                                     \n",
            "                                                                                                  \n",
            " block_12_depthwise_relu (ReLU)  (None, 3, 3, 192)   0           ['block_12_depthwise_BN[0][0]']  \n",
            "                                                                                                  \n",
            " block_12_project (Conv2D)      (None, 3, 3, 32)     6144        ['block_12_depthwise_relu[0][0]']\n",
            "                                                                                                  \n",
            " block_12_project_BN (BatchNorm  (None, 3, 3, 32)    128         ['block_12_project[0][0]']       \n",
            " alization)                                                                                       \n",
            "                                                                                                  \n",
            " block_12_add (Add)             (None, 3, 3, 32)     0           ['block_11_add[0][0]',           \n",
            "                                                                  'block_12_project_BN[0][0]']    \n",
            "                                                                                                  \n",
            " block_13_expand (Conv2D)       (None, 3, 3, 192)    6144        ['block_12_add[0][0]']           \n",
            "                                                                                                  \n",
            " block_13_expand_BN (BatchNorma  (None, 3, 3, 192)   768         ['block_13_expand[0][0]']        \n",
            " lization)                                                                                        \n",
            "                                                                                                  \n",
            " block_13_expand_relu (ReLU)    (None, 3, 3, 192)    0           ['block_13_expand_BN[0][0]']     \n",
            "                                                                                                  \n",
            " block_13_pad (ZeroPadding2D)   (None, 5, 5, 192)    0           ['block_13_expand_relu[0][0]']   \n",
            "                                                                                                  \n",
            " block_13_depthwise (DepthwiseC  (None, 2, 2, 192)   1728        ['block_13_pad[0][0]']           \n",
            " onv2D)                                                                                           \n",
            "                                                                                                  \n",
            " block_13_depthwise_BN (BatchNo  (None, 2, 2, 192)   768         ['block_13_depthwise[0][0]']     \n",
            " rmalization)                                                                                     \n",
            "                                                                                                  \n",
            " block_13_depthwise_relu (ReLU)  (None, 2, 2, 192)   0           ['block_13_depthwise_BN[0][0]']  \n",
            "                                                                                                  \n",
            " block_13_project (Conv2D)      (None, 2, 2, 56)     10752       ['block_13_depthwise_relu[0][0]']\n",
            "                                                                                                  \n",
            " block_13_project_BN (BatchNorm  (None, 2, 2, 56)    224         ['block_13_project[0][0]']       \n",
            " alization)                                                                                       \n",
            "                                                                                                  \n",
            " block_14_expand (Conv2D)       (None, 2, 2, 336)    18816       ['block_13_project_BN[0][0]']    \n",
            "                                                                                                  \n",
            " block_14_expand_BN (BatchNorma  (None, 2, 2, 336)   1344        ['block_14_expand[0][0]']        \n",
            " lization)                                                                                        \n",
            "                                                                                                  \n",
            " block_14_expand_relu (ReLU)    (None, 2, 2, 336)    0           ['block_14_expand_BN[0][0]']     \n",
            "                                                                                                  \n",
            " block_14_depthwise (DepthwiseC  (None, 2, 2, 336)   3024        ['block_14_expand_relu[0][0]']   \n",
            " onv2D)                                                                                           \n",
            "                                                                                                  \n",
            " block_14_depthwise_BN (BatchNo  (None, 2, 2, 336)   1344        ['block_14_depthwise[0][0]']     \n",
            " rmalization)                                                                                     \n",
            "                                                                                                  \n",
            " block_14_depthwise_relu (ReLU)  (None, 2, 2, 336)   0           ['block_14_depthwise_BN[0][0]']  \n",
            "                                                                                                  \n",
            " block_14_project (Conv2D)      (None, 2, 2, 56)     18816       ['block_14_depthwise_relu[0][0]']\n",
            "                                                                                                  \n",
            " block_14_project_BN (BatchNorm  (None, 2, 2, 56)    224         ['block_14_project[0][0]']       \n",
            " alization)                                                                                       \n",
            "                                                                                                  \n",
            " block_14_add (Add)             (None, 2, 2, 56)     0           ['block_13_project_BN[0][0]',    \n",
            "                                                                  'block_14_project_BN[0][0]']    \n",
            "                                                                                                  \n",
            " block_15_expand (Conv2D)       (None, 2, 2, 336)    18816       ['block_14_add[0][0]']           \n",
            "                                                                                                  \n",
            " block_15_expand_BN (BatchNorma  (None, 2, 2, 336)   1344        ['block_15_expand[0][0]']        \n",
            " lization)                                                                                        \n",
            "                                                                                                  \n",
            " block_15_expand_relu (ReLU)    (None, 2, 2, 336)    0           ['block_15_expand_BN[0][0]']     \n",
            "                                                                                                  \n",
            " block_15_depthwise (DepthwiseC  (None, 2, 2, 336)   3024        ['block_15_expand_relu[0][0]']   \n",
            " onv2D)                                                                                           \n",
            "                                                                                                  \n",
            " block_15_depthwise_BN (BatchNo  (None, 2, 2, 336)   1344        ['block_15_depthwise[0][0]']     \n",
            " rmalization)                                                                                     \n",
            "                                                                                                  \n",
            " block_15_depthwise_relu (ReLU)  (None, 2, 2, 336)   0           ['block_15_depthwise_BN[0][0]']  \n",
            "                                                                                                  \n",
            " block_15_project (Conv2D)      (None, 2, 2, 56)     18816       ['block_15_depthwise_relu[0][0]']\n",
            "                                                                                                  \n",
            " block_15_project_BN (BatchNorm  (None, 2, 2, 56)    224         ['block_15_project[0][0]']       \n",
            " alization)                                                                                       \n",
            "                                                                                                  \n",
            " block_15_add (Add)             (None, 2, 2, 56)     0           ['block_14_add[0][0]',           \n",
            "                                                                  'block_15_project_BN[0][0]']    \n",
            "                                                                                                  \n",
            " block_16_expand (Conv2D)       (None, 2, 2, 336)    18816       ['block_15_add[0][0]']           \n",
            "                                                                                                  \n",
            " block_16_expand_BN (BatchNorma  (None, 2, 2, 336)   1344        ['block_16_expand[0][0]']        \n",
            " lization)                                                                                        \n",
            "                                                                                                  \n",
            " block_16_expand_relu (ReLU)    (None, 2, 2, 336)    0           ['block_16_expand_BN[0][0]']     \n",
            "                                                                                                  \n",
            " block_16_depthwise (DepthwiseC  (None, 2, 2, 336)   3024        ['block_16_expand_relu[0][0]']   \n",
            " onv2D)                                                                                           \n",
            "                                                                                                  \n",
            " block_16_depthwise_BN (BatchNo  (None, 2, 2, 336)   1344        ['block_16_depthwise[0][0]']     \n",
            " rmalization)                                                                                     \n",
            "                                                                                                  \n",
            " block_16_depthwise_relu (ReLU)  (None, 2, 2, 336)   0           ['block_16_depthwise_BN[0][0]']  \n",
            "                                                                                                  \n",
            " block_16_project (Conv2D)      (None, 2, 2, 112)    37632       ['block_16_depthwise_relu[0][0]']\n",
            "                                                                                                  \n",
            " block_16_project_BN (BatchNorm  (None, 2, 2, 112)   448         ['block_16_project[0][0]']       \n",
            " alization)                                                                                       \n",
            "                                                                                                  \n",
            " Conv_1 (Conv2D)                (None, 2, 2, 1280)   143360      ['block_16_project_BN[0][0]']    \n",
            "                                                                                                  \n",
            " Conv_1_bn (BatchNormalization)  (None, 2, 2, 1280)  5120        ['Conv_1[0][0]']                 \n",
            "                                                                                                  \n",
            " out_relu (ReLU)                (None, 2, 2, 1280)   0           ['Conv_1_bn[0][0]']              \n",
            "                                                                                                  \n",
            " global_average_pooling2d (Glob  (None, 1280)        0           ['out_relu[0][0]']               \n",
            " alAveragePooling2D)                                                                              \n",
            "                                                                                                  \n",
            " dropout (Dropout)              (None, 1280)         0           ['global_average_pooling2d[0][0]'\n",
            "                                                                 ]                                \n",
            "                                                                                                  \n",
            " dense (Dense)                  (None, 2)            2562        ['dropout[0][0]']                \n",
            "                                                                                                  \n",
            "==================================================================================================\n",
            "Total params: 412,770\n",
            "Trainable params: 2,562\n",
            "Non-trainable params: 410,208\n",
            "__________________________________________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "model.fit(\n",
        "  train_ds,\n",
        "  validation_data=val_ds,\n",
        "  epochs=100\n",
        ")"
      ],
      "metadata": {
        "id": "B_7ZAh2r-IIa"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model.save(TF_MODEL)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "1k0txXWK_csy",
        "outputId": "7ba23d05-1ab7-4658-bb05-0c9e8629f000"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "WARNING:absl:Found untraced functions such as _jit_compiled_convolution_op, _jit_compiled_convolution_op, _jit_compiled_convolution_op, _jit_compiled_convolution_op, _jit_compiled_convolution_op while saving (showing 5 of 53). These functions will not be directly callable after loading.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# KEY: Quantize the TensorFlow model with the TFLite converter!!!"
      ],
      "metadata": {
        "id": "2xa3lD75_UPZ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Key: TFLite Converter settings for converting TensorFlow model to the TFLite\n",
        "# FlatBuffer format (i.e. contigious array of bytes, such as 0x1f form) which \n",
        "# contains the model arhitecture and the weights of the traininable layers \n",
        "# (i.e. in this case the 2,562 weights from the dense layer)\n",
        "converter = tf.lite.TFLiteConverter.from_saved_model(TF_MODEL)\n",
        "converter.representative_dataset = tf.lite.RepresentativeDataset(representative_data_gen)\n",
        "converter.optimizations = [tf.lite.Optimize.DEFAULT]\n",
        "converter.target_spec.supported_ops = [tf.lite.OpsSet.TFLITE_BUILTINS_INT8]\n",
        "converter.inference_input_type = tf.int8\n",
        "tfl_model = converter.convert()"
      ],
      "metadata": {
        "id": "JaPuTllmOlsb"
      },
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "size_tfl_model = len(tfl_model)\n",
        "print(len(tfl_model), \"bytes\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "s4KGgWwCPToc",
        "outputId": "3dadfba7-ebb7-4b3d-9419-480a73f09a66"
      },
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "617384 bytes\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "#Key: use xxd to convert the flatbuffer into a C header file to be imported by Arduino IDE\n",
        "# i.e. converts it from TFlite's flatbuffer into C-byte array for microcontroller\n",
        "# efficiency\n",
        "open(\"model.tflite\", \"wb\").write(tfl_model)\n",
        "!apt-get update && apt-get -qq install xxd\n",
        "!xxd -c 60 -i model.tflite > indoor_scene_recognition.h"
      ],
      "metadata": {
        "id": "mf1v-syzPb45"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Testing TFlite model (quantizied model) on the test data "
      ],
      "metadata": {
        "id": "DSeJzf6tivra"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "with zipfile.ZipFile(\"test_samples.zip\", 'r') as zip_ref:\n",
        "    zip_ref.extractall(\".\")\n",
        "test_dir = \"test_dataset\""
      ],
      "metadata": {
        "id": "_3ReZXOT_QY4"
      },
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def representative_data_gen():\n",
        "  for i_value, o_value in repr_ds.batch(1).take(48):\n",
        "    yield [i_value]\n",
        "\n",
        "test_ds = tf.keras.utils.image_dataset_from_directory(test_dir,\n",
        "                                                      interpolation=\"bilinear\",\n",
        "                                                      image_size=(MODEL_INPUT_WIDTH, MODEL_INPUT_HEIGHT))\n",
        "rescale = tf.keras.layers.Rescaling(1./255, offset= -1)\n",
        "test_ds  = test_ds.map(lambda x, y: (rescale(x), y))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0wiiq6rdOg3Y",
        "outputId": "53fffaef-6907-4ce4-833c-f41b8e49163d"
      },
      "execution_count": 35,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Found 10 files belonging to 2 classes.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Initialize the TFLite interpreter\n",
        "interpreter = tf.lite.Interpreter(model_content=tfl_model)\n",
        "\n",
        "# Allocate the tensors\n",
        "interpreter.allocate_tensors()"
      ],
      "metadata": {
        "id": "HTEtOiFBivRo"
      },
      "execution_count": 36,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# input details of the test data (i.e. un-quantizied images/tensors)\n",
        "i_details = interpreter.get_input_details()[0]\n",
        "# output details of the model (i.e. the classification)\n",
        "o_details = interpreter.get_output_details()[0]\n",
        "# the quantizied information that will be used to map the input values from float domain to int domain\n",
        "i_quant = i_details['quantization_parameters']\n",
        "i_scale = i_quant['scales'][0] # i.e. mapping \n",
        "i_zero_point = i_quant['zero_points'][0] # i.e. offset "
      ],
      "metadata": {
        "id": "e_cCIHXWrLvZ"
      },
      "execution_count": 37,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "test_ds0 = test_ds.unbatch()\n",
        "\n",
        "num_correct_samples = 0\n",
        "num_total_samples   = len(list(test_ds0.batch(1)))\n",
        "\n",
        "for i_value, o_value in test_ds0.batch(1):\n",
        "  i_value = (i_value / i_scale) + i_zero_point\n",
        "  i_value = tf.cast(i_value, dtype=tf.int8)\n",
        "  interpreter.set_tensor(i_details[\"index\"], i_value)\n",
        "  interpreter.invoke()\n",
        "  o_pred = interpreter.get_tensor(o_details[\"index\"])[0]\n",
        "\n",
        "  if np.argmax(o_pred) == o_value:\n",
        "    num_correct_samples += 1\n",
        "\n",
        "print(\"Accuracy:\", num_correct_samples/num_total_samples)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "1ktMJv2L3OBN",
        "outputId": "4261dbc6-ab9f-442e-eb17-33490776b05d"
      },
      "execution_count": 38,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Accuracy: 0.4\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# open(\"model.tflite\", \"wb\").write(tfl_model)\n",
        "# !apt-get update && apt-get -qq install xxd\n",
        "# !xxd -c 60 -i model.tflite > indoor_scene_recognition.h"
      ],
      "metadata": {
        "id": "HdAC7DmerZjn"
      },
      "execution_count": 23,
      "outputs": []
    }
  ]
}