Hanacaraka AI - notebook - improved model.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Hanacaraka AI - improved model.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyOigqvyXxuwunIca1Irzl3V",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/IqbalLx/Hanacaraka-AI/blob/master/Hanacaraka%20AI%20-%20notebook%20-%20improved%20model.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IA1hy-pUuSHJ",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 122
        },
        "outputId": "7e9df9d9-a9bd-44d7-b9a1-9b78ba29d829"
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/gdrive/')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Go to this URL in a browser: https://accounts.google.com/o/oauth2/auth?client_id=947318989803-6bn6qk8qdgf4n4g3pfee6491hc0brc4i.apps.googleusercontent.com&redirect_uri=urn%3aietf%3awg%3aoauth%3a2.0%3aoob&response_type=code&scope=email%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdocs.test%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdrive%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdrive.photos.readonly%20https%3a%2f%2fwww.googleapis.com%2fauth%2fpeopleapi.readonly\n",
            "\n",
            "Enter your authorization code:\n",
            "··········\n",
            "Mounted at /content/gdrive/\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IVqd7GPjMQsM",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import zipfile\n",
        "\n",
        "filename = '/content/gdrive/My Drive/HanacarakaAI/datasetv3.zip'\n",
        "zip_ref = zipfile.ZipFile(filename)\n",
        "zip_ref.extractall()\n",
        "zip_ref.close()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oGxRXx_o58uo",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import os\n",
        "import time\n",
        "import math\n",
        "\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "import matplotlib.image as plt_image\n",
        "plt.style.use('seaborn')\n",
        "\n",
        "import tensorflow as tf\n",
        "\n",
        "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
        "from tensorflow.keras.preprocessing import image as keras_image\n",
        "\n",
        "from tensorflow.keras.models import Sequential\n",
        "from tensorflow.keras.layers import Conv2D, MaxPool2D, GlobalAveragePooling2D\n",
        "from tensorflow.keras.layers import Flatten, Dense, Dropout\n",
        "from tensorflow.keras.layers import BatchNormalization\n",
        "\n",
        "from tensorflow.keras.regularizers import l2\n",
        "\n",
        "from tensorflow.keras.optimizers import Adam\n",
        "\n",
        "from tensorflow.keras.callbacks import TensorBoard\n",
        "%load_ext tensorboard"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "n2-Jwsa0Beqy",
        "colab_type": "code",
        "cellView": "form",
        "colab": {}
      },
      "source": [
        "#@title Helper function for plotting Accuracy and Loss\n",
        "def plot(hist):\n",
        "  history = hist.history\n",
        "  history['epoch'] = hist.epoch\n",
        "\n",
        "  plt.figure(figsize=(12, 5))\n",
        "\n",
        "  plt.subplot(121)\n",
        "  plt.plot(history['epoch'], history['loss'], label='Loss')\n",
        "  plt.plot(history['epoch'], history['val_loss'], label='Val Loss', color='orange')\n",
        "  plt.legend()\n",
        "\n",
        "  plt.subplot(122)\n",
        "  plt.plot(history['epoch'], history['accuracy'], label='Acc')\n",
        "  plt.plot(history['epoch'], history['val_accuracy'], label='Val Acc', color='orange')\n",
        "  plt.legend()\n",
        "\n",
        "  return plt.show()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eIuUPT9NlfiR",
        "colab_type": "code",
        "cellView": "both",
        "colab": {}
      },
      "source": [
        "#@title Helper function for testing model\n",
        "classes = ['ba', 'ca', 'da', 'dha', 'ga', 'ha', 'ja', 'ka', 'la', 'ma',\n",
        "           'na', 'nga', 'nya', 'pa', 'ra', 'sa', 'ta', 'tha', 'wa', 'ya']\n",
        "\n",
        "def test(model, width):\n",
        "  test_images_paths = os.listdir('dataset/prediction')\n",
        "  for path in test_images_paths:\n",
        "    image_path = os.path.join('dataset/prediction', path)\n",
        "\n",
        "    image = keras_image.load_img(image_path,\n",
        "                                 color_mode='grayscale',\n",
        "                                 target_size=(width, width))\n",
        "    x = keras_image.img_to_array(image)\n",
        "    x /= 255\n",
        "    x = np.expand_dims(x, axis=0)\n",
        "\n",
        "    test_image = np.vstack([x])\n",
        "    result = model.predict(test_image, batch_size=8)\n",
        "\n",
        "    print(image_path)\n",
        "    print(classes[np.argmax(result)])\n",
        "\n",
        "    preview = plt_image.imread(image_path)\n",
        "    plt.imshow(preview)\n",
        "    plt.show()\n",
        "  return print('Prediction Done')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nQmCyCco55zr",
        "colab_type": "text"
      },
      "source": [
        "# Image Data Generator"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wvIyZ32w6pqr",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        },
        "outputId": "0b7f0e58-5d27-4b98-8783-60d31b52a930"
      },
      "source": [
        "SIZE = 100 #Image width\n",
        "\n",
        "train_datagen = ImageDataGenerator(rescale=1-(1/255),\n",
        "                                   rotation_range=20,\n",
        "                                   zoom_range=0.2,\n",
        "                                   shear_range=0.1)\n",
        "\n",
        "validation_datagen = ImageDataGenerator(rescale=1-(1/255))\n",
        "\n",
        "train_generator = train_datagen.flow_from_directory(\n",
        "    'datasetv2/training',\n",
        "    target_size=(SIZE, SIZE),\n",
        "    batch_size=32,\n",
        "    color_mode='grayscale',\n",
        "    class_mode = 'categorical'\n",
        ")\n",
        "\n",
        "val_generator = validation_datagen.flow_from_directory(\n",
        "    'datasetv2/testing',\n",
        "    target_size=(SIZE, SIZE),\n",
        "    batch_size=32,\n",
        "    color_mode='grayscale',\n",
        "    class_mode = 'categorical'\n",
        ")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found 3120 images belonging to 20 classes.\n",
            "Found 1299 images belonging to 20 classes.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "D-1RJNRXiGIW",
        "colab_type": "text"
      },
      "source": [
        "#Build Improved Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IIvzStHYl3iZ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "reduce_on_plateau = tf.keras.callbacks.ReduceLROnPlateau(\n",
        "                    monitor='val_loss', factor=0.1, patience=5, verbose=0, mode='auto',\n",
        "                    min_delta=0.0001, cooldown=3, min_lr=1e-8\n",
        "                )"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZH7KaknpAiJi",
        "colab_type": "code",
        "cellView": "form",
        "colab": {}
      },
      "source": [
        "#@title Custom callbacks for Early Stopping\n",
        "#I decide to write my own callback class for early stopping\n",
        "#To added some flexibility\n",
        "#Early stop if accuracy already hit more than 80% \n",
        "#and didnt significantly increased after that\n",
        "\n",
        "class early_stop(tf.keras.callbacks.Callback):\n",
        "  def __init__(self, threshold=0.8, patience=3, significance_val=0.001):\n",
        "    super(early_stop, self).__init__()\n",
        "\n",
        "    self.threshold = threshold                                                    #Minimum accuracy\n",
        "    self.patience = patience                                                      #Limit epoch to wait if accuracy no longer increased\n",
        "    self.significance_val = significance_val                                      #Significant value to mark improvement in accuracy\n",
        "    self.best_weights = None                                                      #Best weights record\n",
        "    self.is_reach_threshold = False                                               #Track if already reach threshold value or not\n",
        "    self.best_epoch = 0                                                           #Track best epoch that produce best weights\n",
        "\n",
        "  def on_train_begin(self, logs=None):\n",
        "    self.previous_acc = 0                                                         #Track previous accuracy\n",
        "    self.wait = 0                                                                 #Track epoch when accuracy no longer increased\n",
        "\n",
        "  def on_epoch_end(self, epoch, logs={}):\n",
        "    self.current_acc = logs.get('val_accuracy')                                   #Current epoch's validation accuracy\n",
        "\n",
        "    if self.is_reach_threshold:\n",
        "      if self.current_acc - self.previous_acc > self.significance_val:\n",
        "        self.wait = 0\n",
        "        self.previous_acc = self.current_acc\n",
        "        self.best_weights = self.model.get_weights()\n",
        "        self.best_epoch = epoch\n",
        "        print('\\n[INFO] Best weights updated!')\n",
        "      else:\n",
        "        self.wait += 1\n",
        "        print(f'\\n[INFO] Current wait = {self.wait}')\n",
        "        if self.wait >= self.patience:\n",
        "          self.model.stop_training = True\n",
        "          print(f'\\n[INFO] Validation Accuracy didnt increased after {self.patience} epochs, training stopped...')\n",
        "          self.model.set_weights(self.best_weights)\n",
        "          print(f'\\n[INFO] Restoring best weights from epoch {self.best_epoch + 1} val_acc')\n",
        "    \n",
        "    if self.current_acc > self.threshold and not self.is_reach_threshold:\n",
        "      self.is_reach_threshold = True\n",
        "      self.previous_acc = self.current_acc\n",
        "      self.best_weights = self.model.get_weights()\n",
        "      self.best_epoch = epoch\n",
        "      print('\\n[INFO] Threshold reached! Best weights start here...')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sTUd9h00ZZmR",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 799
        },
        "outputId": "8850fb8c-fab7-4b36-ac1e-27d5c1bbc821"
      },
      "source": [
        "improved_model = Sequential([Conv2D(16, (3, 3), padding='same', kernel_initializer='he_normal', kernel_regularizer=l2(5e-5), activation='relu', input_shape=(SIZE, SIZE, 1)),\n",
        "                             BatchNormalization(),\n",
        "                             MaxPool2D(2, 2),\n",
        "                             Dropout(0.2),\n",
        "                             Conv2D(32, (3, 3), padding='same', kernel_initializer='he_normal', kernel_regularizer=l2(5e-5), activation='relu'),\n",
        "                             BatchNormalization(),\n",
        "                             MaxPool2D(2, 2),\n",
        "                             Dropout(0.2),\n",
        "                             Conv2D(64, (3, 3), padding='same', kernel_initializer='he_normal', kernel_regularizer=l2(5e-5), activation='relu'),\n",
        "                             BatchNormalization(),\n",
        "                             MaxPool2D(2, 2),\n",
        "                             Dropout(0.2),\n",
        "                             Conv2D(128, (3, 3), padding='same', kernel_initializer='he_normal', kernel_regularizer=l2(5e-5), activation='relu'),\n",
        "                             BatchNormalization(),\n",
        "                             MaxPool2D(2, 2),\n",
        "                             Dropout(0.2),\n",
        "                             GlobalAveragePooling2D(),\n",
        "                             Flatten(),\n",
        "                             Dense(20, activation='softmax')])\n",
        "\n",
        "improved_model.compile(\n",
        "    loss='categorical_crossentropy',\n",
        "    optimizer=Adam(1e-2),\n",
        "    metrics=['accuracy']\n",
        ")\n",
        "\n",
        "improved_model.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d (Conv2D)              (None, 100, 100, 16)      160       \n",
            "_________________________________________________________________\n",
            "batch_normalization (BatchNo (None, 100, 100, 16)      64        \n",
            "_________________________________________________________________\n",
            "max_pooling2d (MaxPooling2D) (None, 50, 50, 16)        0         \n",
            "_________________________________________________________________\n",
            "dropout (Dropout)            (None, 50, 50, 16)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_1 (Conv2D)            (None, 50, 50, 32)        4640      \n",
            "_________________________________________________________________\n",
            "batch_normalization_1 (Batch (None, 50, 50, 32)        128       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_1 (MaxPooling2 (None, 25, 25, 32)        0         \n",
            "_________________________________________________________________\n",
            "dropout_1 (Dropout)          (None, 25, 25, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_2 (Conv2D)            (None, 25, 25, 64)        18496     \n",
            "_________________________________________________________________\n",
            "batch_normalization_2 (Batch (None, 25, 25, 64)        256       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_2 (MaxPooling2 (None, 12, 12, 64)        0         \n",
            "_________________________________________________________________\n",
            "dropout_2 (Dropout)          (None, 12, 12, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_3 (Conv2D)            (None, 12, 12, 128)       73856     \n",
            "_________________________________________________________________\n",
            "batch_normalization_3 (Batch (None, 12, 12, 128)       512       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_3 (MaxPooling2 (None, 6, 6, 128)         0         \n",
            "_________________________________________________________________\n",
            "dropout_3 (Dropout)          (None, 6, 6, 128)         0         \n",
            "_________________________________________________________________\n",
            "global_average_pooling2d (Gl (None, 128)               0         \n",
            "_________________________________________________________________\n",
            "flatten (Flatten)            (None, 128)               0         \n",
            "_________________________________________________________________\n",
            "dense (Dense)                (None, 20)                2580      \n",
            "=================================================================\n",
            "Total params: 100,692\n",
            "Trainable params: 100,212\n",
            "Non-trainable params: 480\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Uib1aG6mggzN",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 164
        },
        "outputId": "6f7da373-4217-4763-db74-2cf2e8b8cb73"
      },
      "source": [
        "assert False"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "error",
          "ename": "AssertionError",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mAssertionError\u001b[0m                            Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-36-a871fdc9ebee>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0;32massert\u001b[0m \u001b[0;32mFalse\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
            "\u001b[0;31mAssertionError\u001b[0m: "
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3qPR6Vn7xysJ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "LOGS = '/content/gdrive/My Drive/HanacarakaAI/logs/'\n",
        "NAME = f'improved_model_v2-{time.time()}'\n",
        "tensorboard = TensorBoard(log_dir=os.path.join(LOGS, NAME), histogram_freq=1)\n",
        "\n",
        "early_stop = early_stop(patience=20)\n",
        "\n",
        "%tensorboard --logdir='/content/gdrive/My Drive/HanacarakaAI/logs/'"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wmJUObMkjFFZ",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "d319978d-a6a3-43c5-a0c9-ca30e3bcd7d4"
      },
      "source": [
        "improved_hist = improved_model.fit(\n",
        "    train_generator,\n",
        "    epochs = 100,\n",
        "    steps_per_epoch = int(3120/32),\n",
        "    validation_data = val_generator,\n",
        "    validation_steps= int(1299/32),\n",
        "    callbacks=[reduce_on_plateau, early_stop, tensorboard],\n",
        "    verbose=1\n",
        ")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/100\n",
            "97/97 [==============================] - 27s 278ms/step - loss: 2.7202 - accuracy: 0.2209 - val_loss: 25.3321 - val_accuracy: 0.0508 - lr: 0.0100\n",
            "Epoch 2/100\n",
            "97/97 [==============================] - 27s 273ms/step - loss: 1.4184 - accuracy: 0.5994 - val_loss: 11.1926 - val_accuracy: 0.0500 - lr: 0.0100\n",
            "Epoch 3/100\n",
            "97/97 [==============================] - 26s 266ms/step - loss: 0.8487 - accuracy: 0.7801 - val_loss: 7.6819 - val_accuracy: 0.0734 - lr: 0.0100\n",
            "Epoch 4/100\n",
            "97/97 [==============================] - 26s 268ms/step - loss: 0.7016 - accuracy: 0.8235 - val_loss: 7.3750 - val_accuracy: 0.1266 - lr: 0.0100\n",
            "Epoch 5/100\n",
            "97/97 [==============================] - 26s 267ms/step - loss: 0.6048 - accuracy: 0.8608 - val_loss: 4.4672 - val_accuracy: 0.1477 - lr: 0.0100\n",
            "Epoch 6/100\n",
            "97/97 [==============================] - 26s 266ms/step - loss: 0.5549 - accuracy: 0.8714 - val_loss: 1.3056 - val_accuracy: 0.6578 - lr: 0.0100\n",
            "Epoch 7/100\n",
            "97/97 [==============================] - 26s 267ms/step - loss: 0.5344 - accuracy: 0.8880 - val_loss: 1.7456 - val_accuracy: 0.5344 - lr: 0.0100\n",
            "Epoch 8/100\n",
            "97/97 [==============================] - 26s 265ms/step - loss: 0.5490 - accuracy: 0.8837 - val_loss: 0.9053 - val_accuracy: 0.7594 - lr: 0.0100\n",
            "Epoch 9/100\n",
            "97/97 [==============================] - 26s 266ms/step - loss: 0.5639 - accuracy: 0.8870 - val_loss: 0.9198 - val_accuracy: 0.7727 - lr: 0.0100\n",
            "Epoch 10/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.5134 - accuracy: 0.9032\n",
            "[INFO] Threshold reached! Best weights start here...\n",
            "97/97 [==============================] - 26s 270ms/step - loss: 0.5134 - accuracy: 0.9032 - val_loss: 0.7036 - val_accuracy: 0.8930 - lr: 0.0100\n",
            "Epoch 11/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.4871 - accuracy: 0.9145\n",
            "[INFO] Current wait = 1\n",
            "97/97 [==============================] - 26s 265ms/step - loss: 0.4871 - accuracy: 0.9145 - val_loss: 1.0423 - val_accuracy: 0.7758 - lr: 0.0100\n",
            "Epoch 12/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.5139 - accuracy: 0.9022\n",
            "[INFO] Current wait = 2\n",
            "97/97 [==============================] - 26s 265ms/step - loss: 0.5139 - accuracy: 0.9022 - val_loss: 5.1775 - val_accuracy: 0.5039 - lr: 0.0100\n",
            "Epoch 13/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.4886 - accuracy: 0.9197\n",
            "[INFO] Current wait = 3\n",
            "97/97 [==============================] - 26s 268ms/step - loss: 0.4886 - accuracy: 0.9197 - val_loss: 2.2065 - val_accuracy: 0.5359 - lr: 0.0100\n",
            "Epoch 14/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.4958 - accuracy: 0.9168\n",
            "[INFO] Current wait = 4\n",
            "97/97 [==============================] - 26s 266ms/step - loss: 0.4958 - accuracy: 0.9168 - val_loss: 0.9453 - val_accuracy: 0.8133 - lr: 0.0100\n",
            "Epoch 15/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.5289 - accuracy: 0.9109\n",
            "[INFO] Current wait = 5\n",
            "97/97 [==============================] - 26s 265ms/step - loss: 0.5289 - accuracy: 0.9109 - val_loss: 2.0151 - val_accuracy: 0.6687 - lr: 0.0100\n",
            "Epoch 16/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.4462 - accuracy: 0.9310\n",
            "[INFO] Best weights updated!\n",
            "97/97 [==============================] - 26s 268ms/step - loss: 0.4462 - accuracy: 0.9310 - val_loss: 0.8937 - val_accuracy: 0.8977 - lr: 1.0000e-03\n",
            "Epoch 17/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.3953 - accuracy: 0.9547\n",
            "[INFO] Best weights updated!\n",
            "97/97 [==============================] - 26s 268ms/step - loss: 0.3953 - accuracy: 0.9547 - val_loss: 0.6274 - val_accuracy: 0.9352 - lr: 1.0000e-03\n",
            "Epoch 18/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.3910 - accuracy: 0.9592\n",
            "[INFO] Best weights updated!\n",
            "97/97 [==============================] - 26s 265ms/step - loss: 0.3910 - accuracy: 0.9592 - val_loss: 0.5708 - val_accuracy: 0.9422 - lr: 1.0000e-03\n",
            "Epoch 19/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.3747 - accuracy: 0.9589\n",
            "[INFO] Best weights updated!\n",
            "97/97 [==============================] - 26s 266ms/step - loss: 0.3747 - accuracy: 0.9589 - val_loss: 0.5347 - val_accuracy: 0.9484 - lr: 1.0000e-03\n",
            "Epoch 20/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.3519 - accuracy: 0.9692\n",
            "[INFO] Current wait = 1\n",
            "97/97 [==============================] - 26s 265ms/step - loss: 0.3519 - accuracy: 0.9692 - val_loss: 0.5728 - val_accuracy: 0.9336 - lr: 1.0000e-03\n",
            "Epoch 21/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.3533 - accuracy: 0.9653\n",
            "[INFO] Current wait = 2\n",
            "97/97 [==============================] - 26s 266ms/step - loss: 0.3533 - accuracy: 0.9653 - val_loss: 0.5013 - val_accuracy: 0.9398 - lr: 1.0000e-03\n",
            "Epoch 22/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.3455 - accuracy: 0.9696\n",
            "[INFO] Current wait = 3\n",
            "97/97 [==============================] - 26s 267ms/step - loss: 0.3455 - accuracy: 0.9696 - val_loss: 0.5050 - val_accuracy: 0.9414 - lr: 1.0000e-03\n",
            "Epoch 23/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.3268 - accuracy: 0.9718\n",
            "[INFO] Current wait = 4\n",
            "97/97 [==============================] - 26s 265ms/step - loss: 0.3268 - accuracy: 0.9718 - val_loss: 0.5103 - val_accuracy: 0.9398 - lr: 1.0000e-03\n",
            "Epoch 24/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.3328 - accuracy: 0.9692\n",
            "[INFO] Current wait = 5\n",
            "97/97 [==============================] - 26s 265ms/step - loss: 0.3328 - accuracy: 0.9692 - val_loss: 0.4982 - val_accuracy: 0.9422 - lr: 1.0000e-03\n",
            "Epoch 25/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.3249 - accuracy: 0.9679\n",
            "[INFO] Current wait = 6\n",
            "97/97 [==============================] - 26s 265ms/step - loss: 0.3249 - accuracy: 0.9679 - val_loss: 0.4741 - val_accuracy: 0.9414 - lr: 1.0000e-03\n",
            "Epoch 26/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.3065 - accuracy: 0.9764\n",
            "[INFO] Current wait = 7\n",
            "97/97 [==============================] - 26s 267ms/step - loss: 0.3065 - accuracy: 0.9764 - val_loss: 0.4766 - val_accuracy: 0.9445 - lr: 1.0000e-03\n",
            "Epoch 27/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.3087 - accuracy: 0.9741\n",
            "[INFO] Current wait = 8\n",
            "97/97 [==============================] - 26s 265ms/step - loss: 0.3087 - accuracy: 0.9741 - val_loss: 0.4791 - val_accuracy: 0.9445 - lr: 1.0000e-03\n",
            "Epoch 28/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.3106 - accuracy: 0.9702\n",
            "[INFO] Current wait = 9\n",
            "97/97 [==============================] - 26s 265ms/step - loss: 0.3106 - accuracy: 0.9702 - val_loss: 0.4701 - val_accuracy: 0.9469 - lr: 1.0000e-03\n",
            "Epoch 29/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2935 - accuracy: 0.9754\n",
            "[INFO] Best weights updated!\n",
            "97/97 [==============================] - 26s 265ms/step - loss: 0.2935 - accuracy: 0.9754 - val_loss: 0.4396 - val_accuracy: 0.9500 - lr: 1.0000e-03\n",
            "Epoch 30/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.3131 - accuracy: 0.9670\n",
            "[INFO] Best weights updated!\n",
            "97/97 [==============================] - 25s 261ms/step - loss: 0.3131 - accuracy: 0.9670 - val_loss: 0.3937 - val_accuracy: 0.9602 - lr: 1.0000e-03\n",
            "Epoch 31/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2879 - accuracy: 0.9734\n",
            "[INFO] Current wait = 1\n",
            "97/97 [==============================] - 26s 264ms/step - loss: 0.2879 - accuracy: 0.9734 - val_loss: 0.4487 - val_accuracy: 0.9492 - lr: 1.0000e-03\n",
            "Epoch 32/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2942 - accuracy: 0.9764\n",
            "[INFO] Current wait = 2\n",
            "97/97 [==============================] - 26s 266ms/step - loss: 0.2942 - accuracy: 0.9764 - val_loss: 0.4397 - val_accuracy: 0.9531 - lr: 1.0000e-03\n",
            "Epoch 33/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2791 - accuracy: 0.9760\n",
            "[INFO] Current wait = 3\n",
            "97/97 [==============================] - 26s 264ms/step - loss: 0.2791 - accuracy: 0.9760 - val_loss: 0.3751 - val_accuracy: 0.9578 - lr: 1.0000e-03\n",
            "Epoch 34/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2776 - accuracy: 0.9764\n",
            "[INFO] Current wait = 4\n",
            "97/97 [==============================] - 26s 265ms/step - loss: 0.2776 - accuracy: 0.9764 - val_loss: 0.4094 - val_accuracy: 0.9508 - lr: 1.0000e-03\n",
            "Epoch 35/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2767 - accuracy: 0.9757\n",
            "[INFO] Current wait = 5\n",
            "97/97 [==============================] - 26s 264ms/step - loss: 0.2767 - accuracy: 0.9757 - val_loss: 0.4409 - val_accuracy: 0.9438 - lr: 1.0000e-03\n",
            "Epoch 36/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2570 - accuracy: 0.9822\n",
            "[INFO] Current wait = 6\n",
            "97/97 [==============================] - 25s 261ms/step - loss: 0.2570 - accuracy: 0.9822 - val_loss: 0.4369 - val_accuracy: 0.9461 - lr: 1.0000e-03\n",
            "Epoch 37/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2597 - accuracy: 0.9790\n",
            "[INFO] Current wait = 7\n",
            "97/97 [==============================] - 25s 261ms/step - loss: 0.2597 - accuracy: 0.9790 - val_loss: 0.3642 - val_accuracy: 0.9555 - lr: 1.0000e-03\n",
            "Epoch 38/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2666 - accuracy: 0.9751\n",
            "[INFO] Current wait = 8\n",
            "97/97 [==============================] - 26s 263ms/step - loss: 0.2666 - accuracy: 0.9751 - val_loss: 0.3895 - val_accuracy: 0.9477 - lr: 1.0000e-03\n",
            "Epoch 39/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2557 - accuracy: 0.9783\n",
            "[INFO] Best weights updated!\n",
            "97/97 [==============================] - 25s 261ms/step - loss: 0.2557 - accuracy: 0.9783 - val_loss: 0.3456 - val_accuracy: 0.9617 - lr: 1.0000e-03\n",
            "Epoch 40/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2508 - accuracy: 0.9803\n",
            "[INFO] Current wait = 1\n",
            "97/97 [==============================] - 25s 262ms/step - loss: 0.2508 - accuracy: 0.9803 - val_loss: 0.3630 - val_accuracy: 0.9563 - lr: 1.0000e-03\n",
            "Epoch 41/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2489 - accuracy: 0.9780\n",
            "[INFO] Current wait = 2\n",
            "97/97 [==============================] - 25s 263ms/step - loss: 0.2489 - accuracy: 0.9780 - val_loss: 0.3667 - val_accuracy: 0.9570 - lr: 1.0000e-03\n",
            "Epoch 42/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2401 - accuracy: 0.9799\n",
            "[INFO] Current wait = 3\n",
            "97/97 [==============================] - 25s 262ms/step - loss: 0.2401 - accuracy: 0.9799 - val_loss: 0.4199 - val_accuracy: 0.9461 - lr: 1.0000e-03\n",
            "Epoch 43/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2426 - accuracy: 0.9770\n",
            "[INFO] Current wait = 4\n",
            "97/97 [==============================] - 25s 260ms/step - loss: 0.2426 - accuracy: 0.9770 - val_loss: 0.4526 - val_accuracy: 0.9453 - lr: 1.0000e-03\n",
            "Epoch 44/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2367 - accuracy: 0.9793\n",
            "[INFO] Current wait = 5\n",
            "97/97 [==============================] - 25s 261ms/step - loss: 0.2367 - accuracy: 0.9793 - val_loss: 0.3651 - val_accuracy: 0.9508 - lr: 1.0000e-03\n",
            "Epoch 45/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2291 - accuracy: 0.9828\n",
            "[INFO] Current wait = 6\n",
            "97/97 [==============================] - 26s 266ms/step - loss: 0.2291 - accuracy: 0.9828 - val_loss: 0.3677 - val_accuracy: 0.9531 - lr: 1.0000e-04\n",
            "Epoch 46/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2164 - accuracy: 0.9867\n",
            "[INFO] Current wait = 7\n",
            "97/97 [==============================] - 26s 267ms/step - loss: 0.2164 - accuracy: 0.9867 - val_loss: 0.3686 - val_accuracy: 0.9539 - lr: 1.0000e-04\n",
            "Epoch 47/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2225 - accuracy: 0.9845\n",
            "[INFO] Current wait = 8\n",
            "97/97 [==============================] - 26s 265ms/step - loss: 0.2225 - accuracy: 0.9845 - val_loss: 0.3701 - val_accuracy: 0.9539 - lr: 1.0000e-04\n",
            "Epoch 48/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2248 - accuracy: 0.9809\n",
            "[INFO] Current wait = 9\n",
            "97/97 [==============================] - 26s 266ms/step - loss: 0.2248 - accuracy: 0.9809 - val_loss: 0.3651 - val_accuracy: 0.9547 - lr: 1.0000e-04\n",
            "Epoch 49/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2261 - accuracy: 0.9815\n",
            "[INFO] Current wait = 10\n",
            "97/97 [==============================] - 26s 265ms/step - loss: 0.2261 - accuracy: 0.9815 - val_loss: 0.3754 - val_accuracy: 0.9539 - lr: 1.0000e-04\n",
            "Epoch 50/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2202 - accuracy: 0.9832\n",
            "[INFO] Current wait = 11\n",
            "97/97 [==============================] - 26s 263ms/step - loss: 0.2202 - accuracy: 0.9832 - val_loss: 0.3847 - val_accuracy: 0.9523 - lr: 1.0000e-04\n",
            "Epoch 51/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2285 - accuracy: 0.9800\n",
            "[INFO] Current wait = 12\n",
            "97/97 [==============================] - 26s 265ms/step - loss: 0.2285 - accuracy: 0.9800 - val_loss: 0.3963 - val_accuracy: 0.9523 - lr: 1.0000e-04\n",
            "Epoch 52/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2275 - accuracy: 0.9806\n",
            "[INFO] Current wait = 13\n",
            "97/97 [==============================] - 26s 267ms/step - loss: 0.2275 - accuracy: 0.9806 - val_loss: 0.3855 - val_accuracy: 0.9531 - lr: 1.0000e-05\n",
            "Epoch 53/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2224 - accuracy: 0.9825\n",
            "[INFO] Current wait = 14\n",
            "97/97 [==============================] - 26s 265ms/step - loss: 0.2224 - accuracy: 0.9825 - val_loss: 0.3733 - val_accuracy: 0.9539 - lr: 1.0000e-05\n",
            "Epoch 54/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2221 - accuracy: 0.9832\n",
            "[INFO] Current wait = 15\n",
            "97/97 [==============================] - 26s 268ms/step - loss: 0.2221 - accuracy: 0.9832 - val_loss: 0.3792 - val_accuracy: 0.9539 - lr: 1.0000e-05\n",
            "Epoch 55/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2233 - accuracy: 0.9822\n",
            "[INFO] Current wait = 16\n",
            "97/97 [==============================] - 26s 266ms/step - loss: 0.2233 - accuracy: 0.9822 - val_loss: 0.3789 - val_accuracy: 0.9531 - lr: 1.0000e-05\n",
            "Epoch 56/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2274 - accuracy: 0.9825\n",
            "[INFO] Current wait = 17\n",
            "97/97 [==============================] - 26s 266ms/step - loss: 0.2274 - accuracy: 0.9825 - val_loss: 0.3677 - val_accuracy: 0.9547 - lr: 1.0000e-05\n",
            "Epoch 57/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2199 - accuracy: 0.9835\n",
            "[INFO] Current wait = 18\n",
            "97/97 [==============================] - 26s 268ms/step - loss: 0.2199 - accuracy: 0.9835 - val_loss: 0.3760 - val_accuracy: 0.9539 - lr: 1.0000e-05\n",
            "Epoch 58/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2194 - accuracy: 0.9828\n",
            "[INFO] Current wait = 19\n",
            "97/97 [==============================] - 26s 265ms/step - loss: 0.2194 - accuracy: 0.9828 - val_loss: 0.3709 - val_accuracy: 0.9547 - lr: 1.0000e-05\n",
            "Epoch 59/100\n",
            "97/97 [==============================] - ETA: 0s - loss: 0.2232 - accuracy: 0.9838\n",
            "[INFO] Current wait = 20\n",
            "\n",
            "[INFO] Validation Accuracy didnt increased after 20 epochs, training stopped...\n",
            "\n",
            "[INFO] Restoring best weights from epoch 39 val_acc\n",
            "97/97 [==============================] - 26s 265ms/step - loss: 0.2232 - accuracy: 0.9838 - val_loss: 0.3660 - val_accuracy: 0.9555 - lr: 1.0000e-06\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iqhkE382B5-o",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 320
        },
        "outputId": "9673063e-a1df-4f62-94e6-63e11efa5637"
      },
      "source": [
        "plot(improved_hist)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 864x360 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LS9vj8XCkLRT",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "f2c98566-ddfd-4ba6-8a2b-93c29829ad8d"
      },
      "source": [
        "test(improved_model, SIZE)\n",
        "assert False"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "dataset/prediction/nga220.pred.png\n",
            "nga\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 576x396 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "dataset/prediction/ba49.mine2.png\n",
            "ba\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 576x396 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "dataset/prediction/ra164.pred.png\n",
            "ra\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 576x396 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "dataset/prediction/ma.0.jpg\n",
            "tha\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 576x396 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "dataset/prediction/ma.1.jpg\n",
            "ma\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 576x396 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Prediction Done\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "error",
          "ename": "AssertionError",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mAssertionError\u001b[0m                            Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-40-551909e3ac5b>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0mtest\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mimproved_model\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mSIZE\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 2\u001b[0;31m \u001b[0;32massert\u001b[0m \u001b[0;32mFalse\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
            "\u001b[0;31mAssertionError\u001b[0m: "
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "V9vUG5OHnuRn",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 105
        },
        "outputId": "a66de91c-c5f2-4697-f3d9-839d63a55d3e"
      },
      "source": [
        "MODEL_PATH = '/content/gdrive/My Drive/HanacarakaAI/model/improved/v3'\n",
        "tf.saved_model.save(improved_model, os.path.join(MODEL_PATH, 'saved_model'))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/resource_variable_ops.py:1817: calling BaseResourceVariable.__init__ (from tensorflow.python.ops.resource_variable_ops) with constraint is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "If using Keras pass *_constraint arguments to layers.\n",
            "INFO:tensorflow:Assets written to: /content/gdrive/My Drive/HanacarakaAI/model/improved/v3/saved_model/assets\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "8Lg6XpNWnuR9",
        "colab": {}
      },
      "source": [
        "tf.keras.models.save_model(improved_model, os.path.join(MODEL_PATH, 'improved_model_v3.h5'))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "B3xEVjxQG44u",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}