notebooks/02_plant_village_tomato_modeling.ipynb

{
  "cells": [
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "q0VQpQmo3V3q"
      },
      "outputs": [],
      "source": [
        "# imports\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "import os\n",
        "import pandas as pd\n",
        "import seaborn as sns\n",
        "\n",
        "from tensorflow.keras.layers import Dense, Dropout, Flatten, Conv2D, MaxPooling2D\n",
        "from tensorflow.keras.regularizers import l2\n",
        "from tensorflow.keras.callbacks import EarlyStopping\n",
        "from tensorflow.keras.models import Sequential\n",
        "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
        "from tensorflow.keras import layers\n",
        "\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.metrics import confusion_matrix, classification_report"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "np.random.seed(1)\n",
        "from tensorflow.keras.utils import set_random_seed\n",
        "set_random_seed(1)"
      ],
      "metadata": {
        "id": "uFCxahI0InHK"
      },
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4wiAqVWn3w2e"
      },
      "source": []
    },
    {
      "cell_type": "code",
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "pi1huuqLQvaL",
        "outputId": "06f32a9d-85dc-4291-ef18-63aa2b6a2d1c"
      },
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mounted at /content/drive\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "id": "ByqYzKd-4ohi"
      },
      "outputs": [],
      "source": [
        "!unzip -q /content/drive/MyDrive/datasets/Tomato.zip"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "NFrXyUR-3eVK",
        "outputId": "67144cbc-b667-4487-aa57-6d8e7f8670a8"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Found 3601 images belonging to 3 classes.\n",
            "Found 899 images belonging to 3 classes.\n"
          ]
        }
      ],
      "source": [
        "# initialize the data generator\n",
        "datagen = ImageDataGenerator(validation_split=0.2)\n",
        "\n",
        "# get training data\n",
        "train_ds = datagen.flow_from_directory(\n",
        "    '/content/Tomato',\n",
        "    target_size=(256, 256),\n",
        "    batch_size=32,\n",
        "    class_mode='categorical',\n",
        "    subset='training',  # set as training data\n",
        "    seed=42\n",
        ")\n",
        "\n",
        "# get validation data\n",
        "val_ds = datagen.flow_from_directory(\n",
        "    '/content/Tomato',\n",
        "    target_size=(256, 256),\n",
        "    batch_size=32,\n",
        "    class_mode='categorical',\n",
        "    subset='validation',  # set as validation data\n",
        "    seed=42,\n",
        "    shuffle=False\n",
        ")"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "train_ds.class_indices"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "QwSmOQIA_ARs",
        "outputId": "238abde3-76e5-4260-c39b-f518d8278fc6"
      },
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{'Tomato_Early_blight': 0, 'Tomato_Late_blight': 1, 'Tomato_healthy': 2}"
            ]
          },
          "metadata": {},
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {
        "id": "32wHG4bY30mT"
      },
      "outputs": [],
      "source": [
        "es = EarlyStopping(patience = 5)\n",
        "\n",
        "model1 = Sequential()\n",
        "\n",
        "model1.add(layers.Input(shape=(256,256,3)))\n",
        "model1.add(layers.Rescaling(1./255))\n",
        "\n",
        "# Convolutional and pooling layers\n",
        "model1.add(Conv2D(64, (3,3), activation='relu'))\n",
        "model1.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "model1.add(Conv2D(64, (3,3), activation='relu'))\n",
        "model1.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "model1.add(Flatten())\n",
        "\n",
        "# Dense Layers\n",
        "model1.add(Dense(64, activation='relu'))\n",
        "\n",
        "# Output\n",
        "model1.add(Dense(3, activation='softmax'))\n",
        "\n",
        "model1.compile(\n",
        "    optimizer=\"adam\",\n",
        "    loss=\"categorical_crossentropy\",\n",
        "    metrics=['accuracy']\n",
        ")"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "model1.input_shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "MLCzvvruR__K",
        "outputId": "0dd486b5-c942-432c-ed11-4da70c5b118d"
      },
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(None, 256, 256, 3)"
            ]
          },
          "metadata": {},
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "model1.summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "4ZBd-PxxSBc_",
        "outputId": "80398a21-92d5-46c4-8fc6-88c37eab16ad"
      },
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " rescaling (Rescaling)       (None, 256, 256, 3)       0         \n",
            "                                                                 \n",
            " conv2d (Conv2D)             (None, 254, 254, 64)      1792      \n",
            "                                                                 \n",
            " max_pooling2d (MaxPooling2D  (None, 127, 127, 64)     0         \n",
            " )                                                               \n",
            "                                                                 \n",
            " conv2d_1 (Conv2D)           (None, 125, 125, 64)      36928     \n",
            "                                                                 \n",
            " max_pooling2d_1 (MaxPooling  (None, 62, 62, 64)       0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " flatten (Flatten)           (None, 246016)            0         \n",
            "                                                                 \n",
            " dense (Dense)               (None, 64)                15745088  \n",
            "                                                                 \n",
            " dense_1 (Dense)             (None, 3)                 195       \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 15,784,003\n",
            "Trainable params: 15,784,003\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 10,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "1euSFn2q3-Pk",
        "outputId": "b98db080-2c76-42c6-e5f8-d95c6e5e56b2"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "113/113 [==============================] - 28s 140ms/step - loss: 0.8326 - accuracy: 0.7134 - val_loss: 0.3233 - val_accuracy: 0.8565\n",
            "Epoch 2/20\n",
            "113/113 [==============================] - 13s 118ms/step - loss: 0.2582 - accuracy: 0.8936 - val_loss: 0.2789 - val_accuracy: 0.8832\n",
            "Epoch 3/20\n",
            "113/113 [==============================] - 14s 126ms/step - loss: 0.1591 - accuracy: 0.9392 - val_loss: 0.4511 - val_accuracy: 0.8509\n",
            "Epoch 4/20\n",
            "113/113 [==============================] - 14s 124ms/step - loss: 0.0946 - accuracy: 0.9661 - val_loss: 0.2550 - val_accuracy: 0.9066\n",
            "Epoch 5/20\n",
            "113/113 [==============================] - 14s 123ms/step - loss: 0.0373 - accuracy: 0.9883 - val_loss: 0.2899 - val_accuracy: 0.9121\n",
            "Epoch 6/20\n",
            "113/113 [==============================] - 13s 117ms/step - loss: 0.0333 - accuracy: 0.9903 - val_loss: 0.2642 - val_accuracy: 0.9210\n",
            "Epoch 7/20\n",
            "113/113 [==============================] - 13s 113ms/step - loss: 0.0452 - accuracy: 0.9847 - val_loss: 0.2845 - val_accuracy: 0.9110\n",
            "Epoch 8/20\n",
            "113/113 [==============================] - 17s 146ms/step - loss: 0.0194 - accuracy: 0.9936 - val_loss: 0.4648 - val_accuracy: 0.8932\n",
            "Epoch 9/20\n",
            "113/113 [==============================] - 14s 122ms/step - loss: 0.0070 - accuracy: 0.9981 - val_loss: 0.3901 - val_accuracy: 0.9199\n"
          ]
        }
      ],
      "source": [
        "h = model1.fit(train_ds,\n",
        "               validation_data=val_ds,\n",
        "               callbacks = [es],\n",
        "               epochs=20)"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Prompted CHAT-GPT 4 to help grab the labels information from the validation dataset and get it into a numpy array, so that we can use that to make a confusion matrix."
      ],
      "metadata": {
        "id": "Bpyo6fADFqli"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def model_preds(model):\n",
        "  '''\n",
        "  This function will return the y_true_classes and y_pred_clasess\n",
        "  Which will allows us to build a confusion matrix\n",
        "  '''\n",
        "  \n",
        "  y_true_classes = val_ds.labels\n",
        "\n",
        "  y_pred_classes = model.predict(val_ds)\n",
        "\n",
        "  y_pred_classes = np.argmax(y_pred_classes, axis=1)\n",
        "\n",
        "  return (y_true_classes, y_pred_classes)"
      ],
      "metadata": {
        "id": "-aZUs9PKZrOX"
      },
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "y_true_classes, y_pred_classes = model_preds(model1)\n",
        "cm = confusion_matrix(y_true_classes, y_pred_classes)\n",
        "\n",
        "fig, ax = plt.subplots(figsize=(12, 5))\n",
        "\n",
        "sns.heatmap(cm, annot=True, fmt='d')\n",
        "\n",
        "ax.set_yticklabels(['Early Blight', 'Late Blight', 'Healthy'], rotation=0, va=\"center\")\n",
        "ax.set_xticklabels(['Early Blight', 'Late Blight', 'Healthy'], rotation=0, ha=\"center\")\n",
        "\n",
        "plt.xlabel('Predicted')\n",
        "plt.ylabel('Truth')\n",
        "plt.show()"
      ],
      "metadata": {
        "id": "dD2PjeaA_DXU",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 483
        },
        "outputId": "e1952937-19f6-4385-d65f-7f754456b8a5"
      },
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "29/29 [==============================] - 2s 74ms/step\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1200x500 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "report = classification_report(y_true_classes, y_pred_classes, target_names=val_ds.class_indices.keys())\n",
        "print(report)"
      ],
      "metadata": {
        "id": "sj4OuYi0_DeI",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "6eb9a795-decf-4041-d08d-913fb8cb179c"
      },
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "                     precision    recall  f1-score   support\n",
            "\n",
            "Tomato_Early_blight       0.87      0.78      0.82       200\n",
            " Tomato_Late_blight       0.90      0.93      0.91       381\n",
            "     Tomato_healthy       0.97      1.00      0.99       318\n",
            "\n",
            "           accuracy                           0.92       899\n",
            "          macro avg       0.91      0.90      0.91       899\n",
            "       weighted avg       0.92      0.92      0.92       899\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "es = EarlyStopping(patience = 5)\n",
        "\n",
        "model2 = Sequential()\n",
        "\n",
        "model2.add(layers.Input(shape=(256,256,3)))\n",
        "model2.add(layers.Rescaling(1./255))\n",
        "\n",
        "# Convolutional and pooling layers\n",
        "model2.add(Conv2D(64, (3,3), activation='relu'))\n",
        "model2.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "model2.add(Conv2D(64, (3,3), activation='relu'))\n",
        "model2.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "model2.add(Flatten())\n",
        "\n",
        "# Dense Layers\n",
        "model2.add(Dense(32, activation='relu'))\n",
        "model2.add(Dense(32, activation='relu'))\n",
        "model2.add(Dense(32, activation='relu'))\n",
        "\n",
        "# Output\n",
        "model2.add(Dense(3, activation='softmax'))\n",
        "\n",
        "model2.compile(\n",
        "    optimizer=\"adam\",\n",
        "    loss=\"categorical_crossentropy\",\n",
        "    metrics=['accuracy']\n",
        ")\n",
        "h = model2.fit(train_ds,\n",
        "               validation_data=val_ds,\n",
        "               callbacks = [es],\n",
        "               epochs=20)"
      ],
      "metadata": {
        "id": "Y59cUr7w_BL9",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "63ac305d-ed08-4c8f-9ff3-ce1ccb416301"
      },
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "113/113 [==============================] - 16s 123ms/step - loss: 0.6800 - accuracy: 0.6812 - val_loss: 0.3193 - val_accuracy: 0.8521\n",
            "Epoch 2/20\n",
            "113/113 [==============================] - 15s 135ms/step - loss: 0.3103 - accuracy: 0.8698 - val_loss: 0.2645 - val_accuracy: 0.8854\n",
            "Epoch 3/20\n",
            "113/113 [==============================] - 14s 119ms/step - loss: 0.2294 - accuracy: 0.9072 - val_loss: 0.2461 - val_accuracy: 0.8932\n",
            "Epoch 4/20\n",
            "113/113 [==============================] - 13s 116ms/step - loss: 0.1538 - accuracy: 0.9464 - val_loss: 0.2289 - val_accuracy: 0.9010\n",
            "Epoch 5/20\n",
            "113/113 [==============================] - 14s 122ms/step - loss: 0.1237 - accuracy: 0.9550 - val_loss: 0.2354 - val_accuracy: 0.9166\n",
            "Epoch 6/20\n",
            "113/113 [==============================] - 13s 116ms/step - loss: 0.0502 - accuracy: 0.9825 - val_loss: 0.2376 - val_accuracy: 0.9299\n",
            "Epoch 7/20\n",
            "113/113 [==============================] - 14s 119ms/step - loss: 0.0347 - accuracy: 0.9917 - val_loss: 0.2351 - val_accuracy: 0.9188\n",
            "Epoch 8/20\n",
            "113/113 [==============================] - 14s 120ms/step - loss: 0.1369 - accuracy: 0.9528 - val_loss: 0.2960 - val_accuracy: 0.9143\n",
            "Epoch 9/20\n",
            "113/113 [==============================] - 13s 119ms/step - loss: 0.0355 - accuracy: 0.9864 - val_loss: 0.3251 - val_accuracy: 0.9143\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Gets overfit so let's try some regularization."
      ],
      "metadata": {
        "id": "ltb9AQg7t8ld"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "es = EarlyStopping(patience = 5)\n",
        "\n",
        "model3 = Sequential()\n",
        "\n",
        "model3.add(layers.Input(shape=(256,256,3)))\n",
        "model3.add(layers.Rescaling(1./255))\n",
        "\n",
        "# Convolutional and pooling layers\n",
        "model3.add(Conv2D(64, (3,3), activation='relu'))\n",
        "model3.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "model3.add(Conv2D(64, (3,3), activation='relu'))\n",
        "model3.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "model3.add(Flatten())\n",
        "\n",
        "# Dense Layers\n",
        "model3.add(Dense(32, activation='relu'))\n",
        "model3.add(Dense(32, activation='relu'))\n",
        "model3.add(Dense(32, activation='relu'))\n",
        "model3.add(Dropout(0.5))\n",
        "\n",
        "# Output\n",
        "model3.add(Dense(3, activation='softmax'))\n",
        "\n",
        "model3.compile(\n",
        "    optimizer=\"adam\",\n",
        "    loss=\"categorical_crossentropy\",\n",
        "    metrics=['accuracy']\n",
        ")\n",
        "h = model3.fit(train_ds,\n",
        "               validation_data=val_ds,\n",
        "               callbacks = [es],\n",
        "               epochs=20)"
      ],
      "metadata": {
        "id": "zgGwfLlprS1x",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "395c5ee3-8290-4b52-b5ad-ab02637cd1cd"
      },
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "113/113 [==============================] - 16s 122ms/step - loss: 0.9690 - accuracy: 0.5096 - val_loss: 0.6675 - val_accuracy: 0.6785\n",
            "Epoch 2/20\n",
            "113/113 [==============================] - 13s 117ms/step - loss: 0.6218 - accuracy: 0.6920 - val_loss: 0.5270 - val_accuracy: 0.7219\n",
            "Epoch 3/20\n",
            "113/113 [==============================] - 13s 118ms/step - loss: 0.5310 - accuracy: 0.7387 - val_loss: 0.4053 - val_accuracy: 0.7920\n",
            "Epoch 4/20\n",
            "113/113 [==============================] - 14s 122ms/step - loss: 0.4648 - accuracy: 0.7587 - val_loss: 0.3836 - val_accuracy: 0.7998\n",
            "Epoch 5/20\n",
            "113/113 [==============================] - 14s 120ms/step - loss: 0.4134 - accuracy: 0.8039 - val_loss: 0.2922 - val_accuracy: 0.8665\n",
            "Epoch 6/20\n",
            "113/113 [==============================] - 13s 118ms/step - loss: 0.3553 - accuracy: 0.8412 - val_loss: 0.3263 - val_accuracy: 0.8487\n",
            "Epoch 7/20\n",
            "113/113 [==============================] - 13s 116ms/step - loss: 0.3086 - accuracy: 0.8748 - val_loss: 0.3184 - val_accuracy: 0.8865\n",
            "Epoch 8/20\n",
            "113/113 [==============================] - 13s 118ms/step - loss: 0.2301 - accuracy: 0.9120 - val_loss: 0.2787 - val_accuracy: 0.8910\n",
            "Epoch 9/20\n",
            "113/113 [==============================] - 13s 117ms/step - loss: 0.1762 - accuracy: 0.9322 - val_loss: 0.3378 - val_accuracy: 0.8799\n",
            "Epoch 10/20\n",
            "113/113 [==============================] - 13s 115ms/step - loss: 0.1540 - accuracy: 0.9447 - val_loss: 0.4185 - val_accuracy: 0.8721\n",
            "Epoch 11/20\n",
            "113/113 [==============================] - 13s 116ms/step - loss: 0.1285 - accuracy: 0.9483 - val_loss: 0.4552 - val_accuracy: 0.8799\n",
            "Epoch 12/20\n",
            "113/113 [==============================] - 13s 117ms/step - loss: 0.1147 - accuracy: 0.9553 - val_loss: 0.8446 - val_accuracy: 0.8565\n",
            "Epoch 13/20\n",
            "113/113 [==============================] - 13s 117ms/step - loss: 0.1410 - accuracy: 0.9553 - val_loss: 0.4812 - val_accuracy: 0.8587\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Still over fitting a lot, let's add another dropout layer."
      ],
      "metadata": {
        "id": "1M_HCrRSvi6a"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "es = EarlyStopping(patience = 5)\n",
        "\n",
        "model4 = Sequential()\n",
        "\n",
        "model4.add(layers.Input(shape=(256,256,3)))\n",
        "model4.add(layers.Rescaling(1./255))\n",
        "\n",
        "# Convolutional and pooling layers\n",
        "model4.add(Conv2D(64, (3,3), activation='relu'))\n",
        "model4.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "model4.add(Conv2D(64, (3,3), activation='relu'))\n",
        "model4.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "model4.add(Flatten())\n",
        "\n",
        "# Dense Layers\n",
        "model4.add(Dense(32, activation='relu'))\n",
        "model4.add(Dense(32, activation='relu'))\n",
        "model4.add(Dropout(0.5))\n",
        "model4.add(Dense(32, activation='relu'))\n",
        "model4.add(Dropout(0.5))\n",
        "\n",
        "# Output\n",
        "model4.add(Dense(3, activation='softmax'))\n",
        "\n",
        "model4.compile(\n",
        "    optimizer=\"adam\",\n",
        "    loss=\"categorical_crossentropy\",\n",
        "    metrics=['accuracy']\n",
        ")\n",
        "h = model4.fit(train_ds,\n",
        "               validation_data=val_ds,\n",
        "               callbacks = [es],\n",
        "               epochs=20)"
      ],
      "metadata": {
        "id": "LiA4joACu2-l",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "4578fe2c-98b9-476b-a637-6b447227ee66"
      },
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "113/113 [==============================] - 15s 119ms/step - loss: 1.0695 - accuracy: 0.4790 - val_loss: 0.9354 - val_accuracy: 0.6174\n",
            "Epoch 2/20\n",
            "113/113 [==============================] - 13s 113ms/step - loss: 0.7758 - accuracy: 0.7001 - val_loss: 0.9579 - val_accuracy: 0.5684\n",
            "Epoch 3/20\n",
            "113/113 [==============================] - 13s 114ms/step - loss: 0.6397 - accuracy: 0.7481 - val_loss: 0.6983 - val_accuracy: 0.7386\n",
            "Epoch 4/20\n",
            "113/113 [==============================] - 19s 168ms/step - loss: 0.5805 - accuracy: 0.7495 - val_loss: 0.4897 - val_accuracy: 0.7664\n",
            "Epoch 5/20\n",
            "113/113 [==============================] - 13s 115ms/step - loss: 0.5724 - accuracy: 0.7445 - val_loss: 0.4894 - val_accuracy: 0.7709\n",
            "Epoch 6/20\n",
            "113/113 [==============================] - 13s 116ms/step - loss: 0.5153 - accuracy: 0.7565 - val_loss: 0.4738 - val_accuracy: 0.7697\n",
            "Epoch 7/20\n",
            "113/113 [==============================] - 14s 121ms/step - loss: 0.5139 - accuracy: 0.7562 - val_loss: 0.4519 - val_accuracy: 0.7686\n",
            "Epoch 8/20\n",
            "113/113 [==============================] - 14s 122ms/step - loss: 0.4810 - accuracy: 0.7634 - val_loss: 0.4569 - val_accuracy: 0.7675\n",
            "Epoch 9/20\n",
            "113/113 [==============================] - 14s 123ms/step - loss: 0.5033 - accuracy: 0.7587 - val_loss: 0.4068 - val_accuracy: 0.7720\n",
            "Epoch 10/20\n",
            "113/113 [==============================] - 14s 123ms/step - loss: 0.4712 - accuracy: 0.7590 - val_loss: 0.4319 - val_accuracy: 0.7731\n",
            "Epoch 11/20\n",
            "113/113 [==============================] - 13s 119ms/step - loss: 0.4437 - accuracy: 0.7648 - val_loss: 0.3933 - val_accuracy: 0.7597\n",
            "Epoch 12/20\n",
            "113/113 [==============================] - 13s 118ms/step - loss: 0.3944 - accuracy: 0.7706 - val_loss: 0.3399 - val_accuracy: 0.7709\n",
            "Epoch 13/20\n",
            "113/113 [==============================] - 14s 122ms/step - loss: 0.3837 - accuracy: 0.7706 - val_loss: 0.3230 - val_accuracy: 0.7742\n",
            "Epoch 14/20\n",
            "113/113 [==============================] - 13s 113ms/step - loss: 0.4531 - accuracy: 0.7570 - val_loss: 0.3901 - val_accuracy: 0.7686\n",
            "Epoch 15/20\n",
            "113/113 [==============================] - 14s 121ms/step - loss: 0.3825 - accuracy: 0.7670 - val_loss: 0.3602 - val_accuracy: 0.7720\n",
            "Epoch 16/20\n",
            "113/113 [==============================] - 13s 118ms/step - loss: 0.3438 - accuracy: 0.7734 - val_loss: 0.4068 - val_accuracy: 0.7620\n",
            "Epoch 17/20\n",
            "113/113 [==============================] - 15s 136ms/step - loss: 0.3575 - accuracy: 0.7776 - val_loss: 0.3233 - val_accuracy: 0.7742\n",
            "Epoch 18/20\n",
            "113/113 [==============================] - 13s 118ms/step - loss: 0.3423 - accuracy: 0.7762 - val_loss: 0.3237 - val_accuracy: 0.7920\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Still over fitting by a lot. Before trying another dropout layer, let's add in a l2 regularization, and see what the results are."
      ],
      "metadata": {
        "id": "-SfuCKwtwcEA"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "es = EarlyStopping(patience = 5)\n",
        "\n",
        "model5 = Sequential()\n",
        "\n",
        "model5.add(layers.Input(shape=(256,256,3)))\n",
        "model5.add(layers.Rescaling(1./255))\n",
        "\n",
        "# Convolutional and pooling layers\n",
        "model5.add(Conv2D(64, (3,3), activation='relu'))\n",
        "model5.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "model5.add(Conv2D(64, (3,3), activation='relu'))\n",
        "model5.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "model5.add(Flatten())\n",
        "\n",
        "# Dense Layers\n",
        "model5.add(Dense(32, activation='relu', kernel_regularizer=l2(0.005)))\n",
        "model5.add(Dense(32, activation='relu', kernel_regularizer=l2(0.005)))\n",
        "model5.add(Dropout(0.5))\n",
        "model5.add(Dense(32, activation='relu', kernel_regularizer=l2(0.005)))\n",
        "model5.add(Dropout(0.5))\n",
        "\n",
        "# Output\n",
        "model5.add(Dense(3, activation='softmax'))\n",
        "\n",
        "model5.compile(\n",
        "    optimizer=\"adam\",\n",
        "    loss=\"categorical_crossentropy\",\n",
        "    metrics=['accuracy']\n",
        ")\n",
        "\n",
        "h = model5.fit(train_ds,\n",
        "               validation_data=val_ds,\n",
        "               callbacks = [es],\n",
        "               epochs=20)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Vz8Abaj6vy78",
        "outputId": "eaf47aad-94c1-4652-9a97-62e48dbd5f75"
      },
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "113/113 [==============================] - 16s 121ms/step - loss: 1.4334 - accuracy: 0.5568 - val_loss: 0.8581 - val_accuracy: 0.7241\n",
            "Epoch 2/20\n",
            "113/113 [==============================] - 14s 121ms/step - loss: 0.9160 - accuracy: 0.6934 - val_loss: 0.7121 - val_accuracy: 0.7519\n",
            "Epoch 3/20\n",
            "113/113 [==============================] - 13s 119ms/step - loss: 0.7345 - accuracy: 0.7448 - val_loss: 0.5919 - val_accuracy: 0.7775\n",
            "Epoch 4/20\n",
            "113/113 [==============================] - 13s 115ms/step - loss: 0.6846 - accuracy: 0.7767 - val_loss: 0.8799 - val_accuracy: 0.7230\n",
            "Epoch 5/20\n",
            "113/113 [==============================] - 13s 117ms/step - loss: 0.6267 - accuracy: 0.8139 - val_loss: 0.5484 - val_accuracy: 0.8320\n",
            "Epoch 6/20\n",
            "113/113 [==============================] - 15s 128ms/step - loss: 0.5664 - accuracy: 0.8278 - val_loss: 0.5485 - val_accuracy: 0.8176\n",
            "Epoch 7/20\n",
            "113/113 [==============================] - 14s 122ms/step - loss: 0.5447 - accuracy: 0.8373 - val_loss: 0.5097 - val_accuracy: 0.8465\n",
            "Epoch 8/20\n",
            "113/113 [==============================] - 14s 122ms/step - loss: 0.5421 - accuracy: 0.8373 - val_loss: 0.4576 - val_accuracy: 0.8843\n",
            "Epoch 9/20\n",
            "113/113 [==============================] - 14s 121ms/step - loss: 0.5322 - accuracy: 0.8495 - val_loss: 0.4718 - val_accuracy: 0.8732\n",
            "Epoch 10/20\n",
            "113/113 [==============================] - 13s 116ms/step - loss: 0.5272 - accuracy: 0.8603 - val_loss: 0.4386 - val_accuracy: 0.8888\n",
            "Epoch 11/20\n",
            "113/113 [==============================] - 14s 121ms/step - loss: 0.4726 - accuracy: 0.8761 - val_loss: 0.3934 - val_accuracy: 0.8943\n",
            "Epoch 12/20\n",
            "113/113 [==============================] - 13s 119ms/step - loss: 0.4481 - accuracy: 0.8798 - val_loss: 0.4114 - val_accuracy: 0.8943\n",
            "Epoch 13/20\n",
            "113/113 [==============================] - 14s 120ms/step - loss: 0.4385 - accuracy: 0.8961 - val_loss: 0.4282 - val_accuracy: 0.8999\n",
            "Epoch 14/20\n",
            "113/113 [==============================] - 14s 120ms/step - loss: 0.4437 - accuracy: 0.8934 - val_loss: 0.4156 - val_accuracy: 0.8943\n",
            "Epoch 15/20\n",
            "113/113 [==============================] - 14s 123ms/step - loss: 0.4376 - accuracy: 0.9006 - val_loss: 0.4054 - val_accuracy: 0.9043\n",
            "Epoch 16/20\n",
            "113/113 [==============================] - 14s 123ms/step - loss: 0.4270 - accuracy: 0.9028 - val_loss: 0.3626 - val_accuracy: 0.9132\n",
            "Epoch 17/20\n",
            "113/113 [==============================] - 14s 121ms/step - loss: 0.4445 - accuracy: 0.9042 - val_loss: 0.3731 - val_accuracy: 0.9132\n",
            "Epoch 18/20\n",
            "113/113 [==============================] - 13s 118ms/step - loss: 0.3824 - accuracy: 0.9164 - val_loss: 0.3350 - val_accuracy: 0.9132\n",
            "Epoch 19/20\n",
            "113/113 [==============================] - 13s 118ms/step - loss: 0.3804 - accuracy: 0.9245 - val_loss: 0.3671 - val_accuracy: 0.9099\n",
            "Epoch 20/20\n",
            "113/113 [==============================] - 14s 121ms/step - loss: 0.4430 - accuracy: 0.9086 - val_loss: 0.4005 - val_accuracy: 0.9099\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "The model just so happened to start with a really low score, so it was able to hit 20 epochs, but in attempts to reproduce, the model would stil overfit and early stop."
      ],
      "metadata": {
        "id": "XnSvi5Rp0KyA"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "es = EarlyStopping(patience = 5)\n",
        "\n",
        "model6 = Sequential()\n",
        "\n",
        "model6.add(layers.Input(shape=(256,256,3)))\n",
        "model6.add(layers.Rescaling(1./255))\n",
        "\n",
        "# Convolutional and pooling layers\n",
        "model6.add(Conv2D(64, (3,3), activation='relu'))\n",
        "model6.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "model6.add(Conv2D(64, (3,3), activation='relu'))\n",
        "model6.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "model6.add(Flatten())\n",
        "\n",
        "# Dense Layers\n",
        "model6.add(Dense(16, activation='relu', kernel_regularizer=l2(0.005)))\n",
        "model6.add(Dense(32, activation='relu', kernel_regularizer=l2(0.005)))\n",
        "model6.add(Dropout(0.5))\n",
        "model6.add(Dense(32, activation='relu', kernel_regularizer=l2(0.005)))\n",
        "model6.add(Dropout(0.5))\n",
        "model6.add(Dense(32, activation='relu', kernel_regularizer=l2(0.005)))\n",
        "\n",
        "# Output\n",
        "model6.add(Dense(3, activation='softmax'))\n",
        "\n",
        "model6.compile(\n",
        "    optimizer=\"adam\",\n",
        "    loss=\"categorical_crossentropy\",\n",
        "    metrics=['accuracy']\n",
        ")\n",
        "\n",
        "h = model6.fit(train_ds,\n",
        "               validation_data=val_ds,\n",
        "               callbacks = [es],\n",
        "               epochs=20)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "cZ-5VtJiy_fF",
        "outputId": "6a160b8a-a5dd-471f-b5fd-96428999e6ac"
      },
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "113/113 [==============================] - 16s 120ms/step - loss: 1.5210 - accuracy: 0.4135 - val_loss: 1.3457 - val_accuracy: 0.4238\n",
            "Epoch 2/20\n",
            "113/113 [==============================] - 13s 118ms/step - loss: 1.2881 - accuracy: 0.4224 - val_loss: 1.2286 - val_accuracy: 0.4238\n",
            "Epoch 3/20\n",
            "113/113 [==============================] - 13s 118ms/step - loss: 1.2086 - accuracy: 0.4249 - val_loss: 1.1800 - val_accuracy: 0.4238\n",
            "Epoch 4/20\n",
            "113/113 [==============================] - 14s 120ms/step - loss: 1.1758 - accuracy: 0.4243 - val_loss: 1.1845 - val_accuracy: 0.4238\n",
            "Epoch 5/20\n",
            "113/113 [==============================] - 13s 117ms/step - loss: 1.1668 - accuracy: 0.4243 - val_loss: 1.1462 - val_accuracy: 0.4238\n",
            "Epoch 6/20\n",
            "113/113 [==============================] - 13s 116ms/step - loss: 1.1192 - accuracy: 0.4243 - val_loss: 1.1098 - val_accuracy: 0.4238\n",
            "Epoch 7/20\n",
            "113/113 [==============================] - 13s 119ms/step - loss: 1.1046 - accuracy: 0.4243 - val_loss: 1.1091 - val_accuracy: 0.4238\n",
            "Epoch 8/20\n",
            "113/113 [==============================] - 13s 117ms/step - loss: 1.0929 - accuracy: 0.4243 - val_loss: 1.0867 - val_accuracy: 0.4238\n",
            "Epoch 9/20\n",
            "113/113 [==============================] - 13s 118ms/step - loss: 1.0759 - accuracy: 0.4243 - val_loss: 1.0767 - val_accuracy: 0.4238\n",
            "Epoch 10/20\n",
            "113/113 [==============================] - 13s 117ms/step - loss: 1.0784 - accuracy: 0.4243 - val_loss: 1.0715 - val_accuracy: 0.4238\n",
            "Epoch 11/20\n",
            "113/113 [==============================] - 13s 118ms/step - loss: 1.0622 - accuracy: 0.4243 - val_loss: 1.0651 - val_accuracy: 0.4238\n",
            "Epoch 12/20\n",
            "113/113 [==============================] - 14s 120ms/step - loss: 1.0599 - accuracy: 0.4243 - val_loss: 1.0990 - val_accuracy: 0.4238\n",
            "Epoch 13/20\n",
            "113/113 [==============================] - 14s 119ms/step - loss: 1.0917 - accuracy: 0.4243 - val_loss: 1.0817 - val_accuracy: 0.4238\n",
            "Epoch 14/20\n",
            "113/113 [==============================] - 14s 121ms/step - loss: 1.0771 - accuracy: 0.4243 - val_loss: 1.0720 - val_accuracy: 0.4238\n",
            "Epoch 15/20\n",
            "113/113 [==============================] - 14s 125ms/step - loss: 1.0685 - accuracy: 0.4243 - val_loss: 1.0598 - val_accuracy: 0.4238\n",
            "Epoch 16/20\n",
            "113/113 [==============================] - 13s 117ms/step - loss: 1.0504 - accuracy: 0.4243 - val_loss: 1.0548 - val_accuracy: 0.4238\n",
            "Epoch 17/20\n",
            "113/113 [==============================] - 13s 116ms/step - loss: 1.0474 - accuracy: 0.4243 - val_loss: 1.0681 - val_accuracy: 0.4238\n",
            "Epoch 18/20\n",
            "113/113 [==============================] - 13s 117ms/step - loss: 1.0531 - accuracy: 0.4243 - val_loss: 1.0511 - val_accuracy: 0.4238\n",
            "Epoch 19/20\n",
            "113/113 [==============================] - 13s 116ms/step - loss: 1.0451 - accuracy: 0.4243 - val_loss: 1.0516 - val_accuracy: 0.4238\n",
            "Epoch 20/20\n",
            "113/113 [==============================] - 13s 114ms/step - loss: 1.0256 - accuracy: 0.4529 - val_loss: 0.9294 - val_accuracy: 0.5528\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "We've been tackling overfitting, but the accuracy on training is still worse than the validation of the original simple but overfit model. Adding more complexity hasn't really helped us that much. Let's start over, but now mess around with convolutions"
      ],
      "metadata": {
        "id": "cNDvE_A54ng7"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "es = EarlyStopping(patience = 5)\n",
        "\n",
        "model7 = Sequential()\n",
        "\n",
        "model7.add(layers.Input(shape=(256,256,3)))\n",
        "model7.add(layers.Rescaling(1./255))\n",
        "\n",
        "# Convolutional and pooling layers\n",
        "model7.add(Conv2D(64, (4,4), activation='relu'))\n",
        "model7.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "model7.add(Conv2D(64, (4,4), activation='relu'))\n",
        "model7.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "model7.add(Flatten())\n",
        "\n",
        "# Dense Layers\n",
        "model7.add(Dense(64, activation='relu'))\n",
        "\n",
        "# Output\n",
        "model7.add(Dense(3, activation='softmax'))\n",
        "\n",
        "model7.compile(\n",
        "    optimizer=\"adam\",\n",
        "    loss=\"categorical_crossentropy\",\n",
        "    metrics=['accuracy']\n",
        ")\n",
        "\n",
        "h = model7.fit(train_ds,\n",
        "               validation_data=val_ds,\n",
        "               callbacks = [es],\n",
        "               epochs=20)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "1A0yds_Y0pTC",
        "outputId": "9fba1628-7bfd-4c86-eb2c-e213f28ef571"
      },
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "113/113 [==============================] - 19s 135ms/step - loss: 1.1411 - accuracy: 0.5496 - val_loss: 1.0088 - val_accuracy: 0.5495\n",
            "Epoch 2/20\n",
            "113/113 [==============================] - 14s 124ms/step - loss: 0.9707 - accuracy: 0.5451 - val_loss: 0.8317 - val_accuracy: 0.6607\n",
            "Epoch 3/20\n",
            "113/113 [==============================] - 14s 120ms/step - loss: 0.7909 - accuracy: 0.6723 - val_loss: 0.8058 - val_accuracy: 0.6752\n",
            "Epoch 4/20\n",
            "113/113 [==============================] - 14s 121ms/step - loss: 0.5624 - accuracy: 0.7884 - val_loss: 0.6833 - val_accuracy: 0.7253\n",
            "Epoch 5/20\n",
            "113/113 [==============================] - 14s 121ms/step - loss: 0.4959 - accuracy: 0.7976 - val_loss: 0.7954 - val_accuracy: 0.7286\n",
            "Epoch 6/20\n",
            "113/113 [==============================] - 14s 122ms/step - loss: 0.3383 - accuracy: 0.8861 - val_loss: 0.7265 - val_accuracy: 0.7464\n",
            "Epoch 7/20\n",
            "113/113 [==============================] - 14s 121ms/step - loss: 0.2028 - accuracy: 0.9384 - val_loss: 0.8730 - val_accuracy: 0.7375\n",
            "Epoch 8/20\n",
            "113/113 [==============================] - 14s 123ms/step - loss: 0.1003 - accuracy: 0.9731 - val_loss: 0.9929 - val_accuracy: 0.7464\n",
            "Epoch 9/20\n",
            "113/113 [==============================] - 14s 123ms/step - loss: 0.0541 - accuracy: 0.9886 - val_loss: 1.1942 - val_accuracy: 0.7486\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "changing the convolution input matrix from 3 x 3 to 4 x 4 didn't have a great effect. Let's go back to 3 x 3 and instead see what happens if we add a conv layer and remove a conv layer."
      ],
      "metadata": {
        "id": "hyIVt4US6GT6"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "es = EarlyStopping(patience = 5)\n",
        "\n",
        "model8 = Sequential()\n",
        "\n",
        "model8.add(layers.Input(shape=(256,256,3)))\n",
        "model8.add(layers.Rescaling(1./255))\n",
        "\n",
        "# Convolutional and pooling layers\n",
        "model8.add(Conv2D(64, (3,3), activation='relu'))\n",
        "model8.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "model8.add(Conv2D(64, (3,3), activation='relu'))\n",
        "model8.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "model8.add(Conv2D(64, (3,3), activation='relu'))\n",
        "model8.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "model8.add(Flatten())\n",
        "\n",
        "# Dense Layers\n",
        "model8.add(Dense(64, activation='relu'))\n",
        "\n",
        "# Output\n",
        "model8.add(Dense(3, activation='softmax'))\n",
        "\n",
        "model8.compile(\n",
        "    optimizer=\"adam\",\n",
        "    loss=\"categorical_crossentropy\",\n",
        "    metrics=['accuracy']\n",
        ")\n",
        "\n",
        "h = model8.fit(train_ds,\n",
        "               validation_data=val_ds,\n",
        "               callbacks = [es],\n",
        "               epochs=20)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "cPjURya55SSt",
        "outputId": "bba2ce8d-7b4f-45a9-ec03-ebac55a9d9f0"
      },
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "113/113 [==============================] - 16s 120ms/step - loss: 0.7404 - accuracy: 0.6454 - val_loss: 0.3634 - val_accuracy: 0.8331\n",
            "Epoch 2/20\n",
            "113/113 [==============================] - 13s 119ms/step - loss: 0.3217 - accuracy: 0.8534 - val_loss: 0.3490 - val_accuracy: 0.8465\n",
            "Epoch 3/20\n",
            "113/113 [==============================] - 14s 119ms/step - loss: 0.3064 - accuracy: 0.8739 - val_loss: 0.4659 - val_accuracy: 0.8209\n",
            "Epoch 4/20\n",
            "113/113 [==============================] - 14s 125ms/step - loss: 0.2194 - accuracy: 0.9147 - val_loss: 0.1971 - val_accuracy: 0.9143\n",
            "Epoch 5/20\n",
            "113/113 [==============================] - 14s 120ms/step - loss: 0.1754 - accuracy: 0.9317 - val_loss: 0.1937 - val_accuracy: 0.9143\n",
            "Epoch 6/20\n",
            "113/113 [==============================] - 14s 120ms/step - loss: 0.1292 - accuracy: 0.9506 - val_loss: 0.1943 - val_accuracy: 0.9199\n",
            "Epoch 7/20\n",
            "113/113 [==============================] - 13s 118ms/step - loss: 0.0986 - accuracy: 0.9658 - val_loss: 0.1476 - val_accuracy: 0.9466\n",
            "Epoch 8/20\n",
            "113/113 [==============================] - 13s 119ms/step - loss: 0.0822 - accuracy: 0.9692 - val_loss: 0.3599 - val_accuracy: 0.8910\n",
            "Epoch 9/20\n",
            "113/113 [==============================] - 14s 120ms/step - loss: 0.0632 - accuracy: 0.9781 - val_loss: 0.2736 - val_accuracy: 0.9132\n",
            "Epoch 10/20\n",
            "113/113 [==============================] - 14s 121ms/step - loss: 0.0406 - accuracy: 0.9869 - val_loss: 0.2064 - val_accuracy: 0.9388\n",
            "Epoch 11/20\n",
            "113/113 [==============================] - 14s 123ms/step - loss: 0.0870 - accuracy: 0.9681 - val_loss: 0.2513 - val_accuracy: 0.9143\n",
            "Epoch 12/20\n",
            "113/113 [==============================] - 14s 126ms/step - loss: 0.0370 - accuracy: 0.9872 - val_loss: 0.2386 - val_accuracy: 0.9199\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "es = EarlyStopping(patience = 5)\n",
        "\n",
        "model9 = Sequential()\n",
        "\n",
        "model9.add(layers.Input(shape=(256,256,3)))\n",
        "model9.add(layers.Rescaling(1./255))\n",
        "\n",
        "# Convolutional and pooling layers\n",
        "model9.add(Conv2D(64, (3,3), activation='relu'))\n",
        "model9.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "model9.add(Flatten())\n",
        "\n",
        "# Dense Layers\n",
        "model9.add(Dense(64, activation='relu'))\n",
        "\n",
        "# Output\n",
        "model9.add(Dense(3, activation='softmax'))\n",
        "\n",
        "model9.compile(\n",
        "    optimizer=\"adam\",\n",
        "    loss=\"categorical_crossentropy\",\n",
        "    metrics=['accuracy']\n",
        ")\n",
        "\n",
        "h = model9.fit(train_ds,\n",
        "               validation_data=val_ds,\n",
        "               callbacks = [es],\n",
        "               epochs=20)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "B7HTw9sR7WKB",
        "outputId": "f2746e50-92cb-4e4b-c394-a4dc21e011d2"
      },
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "113/113 [==============================] - 17s 136ms/step - loss: 3.6017 - accuracy: 0.6754 - val_loss: 0.4043 - val_accuracy: 0.8420\n",
            "Epoch 2/20\n",
            "113/113 [==============================] - 13s 115ms/step - loss: 0.2823 - accuracy: 0.8942 - val_loss: 0.2558 - val_accuracy: 0.8932\n",
            "Epoch 3/20\n",
            "113/113 [==============================] - 13s 113ms/step - loss: 0.1104 - accuracy: 0.9670 - val_loss: 0.2362 - val_accuracy: 0.8999\n",
            "Epoch 4/20\n",
            "113/113 [==============================] - 13s 113ms/step - loss: 0.0413 - accuracy: 0.9897 - val_loss: 0.2254 - val_accuracy: 0.9155\n",
            "Epoch 5/20\n",
            "113/113 [==============================] - 13s 115ms/step - loss: 0.0210 - accuracy: 0.9956 - val_loss: 0.3152 - val_accuracy: 0.8966\n",
            "Epoch 6/20\n",
            "113/113 [==============================] - 13s 115ms/step - loss: 0.0068 - accuracy: 0.9992 - val_loss: 0.2654 - val_accuracy: 0.9210\n",
            "Epoch 7/20\n",
            "113/113 [==============================] - 13s 114ms/step - loss: 0.0014 - accuracy: 1.0000 - val_loss: 0.2725 - val_accuracy: 0.9177\n",
            "Epoch 8/20\n",
            "113/113 [==============================] - 13s 112ms/step - loss: 7.2649e-04 - accuracy: 1.0000 - val_loss: 0.3033 - val_accuracy: 0.9188\n",
            "Epoch 9/20\n",
            "113/113 [==============================] - 13s 115ms/step - loss: 3.8801e-04 - accuracy: 1.0000 - val_loss: 0.3098 - val_accuracy: 0.9221\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Looks like our training accuracy is much more overfit with less validation accuracy. Currently model 8 seems to have the best performance it has the highest accuracy (96% on training and 92% on validation accuracy) and is less over fit than other models. Our second best model was model 2 with a train accuracy of 99% and a validation accuracy of 92%.\n",
        "\n",
        "Lets see how well they do with a confusion matrix, because of their similar accuacy scores it is most likely they are going to have similar misclassifiations.\n",
        "___________________________"
      ],
      "metadata": {
        "id": "ZaT2Gmq7yeGY"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Confusion matrixs"
      ],
      "metadata": {
        "id": "D0ia5-eJ02JU"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "y_true_classes, y_pred_classes = model_preds(model8)\n",
        "cm = confusion_matrix(y_true_classes, y_pred_classes)\n",
        "\n",
        "fig, ax = plt.subplots(figsize=(12, 5))\n",
        "\n",
        "sns.heatmap(cm, annot=True, fmt='d')\n",
        "\n",
        "ax.set_yticklabels(['Early Blight', 'Late Blight', 'Healthy'], rotation=0, va=\"center\")\n",
        "ax.set_xticklabels(['Early Blight', 'Late Blight', 'Healthy'], rotation=0, ha=\"center\")\n",
        "plt.title('model 8 matrix')\n",
        "plt.xlabel('Predicted')\n",
        "plt.ylabel('Truth')\n",
        "plt.savefig('Model_8_Matrix.png')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 504
        },
        "id": "WlyuQWSN2GJb",
        "outputId": "e1bdfb25-ae5d-4fb0-ee3b-a20d52fb80cf"
      },
      "execution_count": 28,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "29/29 [==============================] - 4s 144ms/step\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1200x500 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "report = classification_report(y_true_classes, y_pred_classes, target_names=val_ds.class_indices.keys())\n",
        "print(report)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Syt-58WG4DQK",
        "outputId": "40b75847-9f76-4ed4-a680-270f35f98074"
      },
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "                     precision    recall  f1-score   support\n",
            "\n",
            "Tomato_Early_blight       0.88      0.77      0.82       200\n",
            " Tomato_Late_blight       0.88      0.94      0.91       381\n",
            "     Tomato_healthy       0.99      0.99      0.99       318\n",
            "\n",
            "           accuracy                           0.92       899\n",
            "          macro avg       0.92      0.90      0.91       899\n",
            "       weighted avg       0.92      0.92      0.92       899\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "y_true_classes, y_pred_classes = model_preds(model2)\n",
        "cm = confusion_matrix(y_true_classes, y_pred_classes)\n",
        "\n",
        "fig, ax = plt.subplots(figsize=(12, 5))\n",
        "\n",
        "sns.heatmap(cm, annot=True, fmt='d')\n",
        "\n",
        "ax.set_yticklabels(['Early Blight', 'Late Blight', 'Healthy'], rotation=0, va=\"center\")\n",
        "ax.set_xticklabels(['Early Blight', 'Late Blight', 'Healthy'], rotation=0, ha=\"center\")\n",
        "plt.title('model 2 matrix')\n",
        "plt.xlabel('Predicted')\n",
        "plt.ylabel('Truth')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 504
        },
        "id": "1B6M9Mjx05QK",
        "outputId": "df32a057-b016-4396-c89b-90c5ce910b26"
      },
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "29/29 [==============================] - 2s 76ms/step\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1200x500 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "report = classification_report(y_true_classes, y_pred_classes, target_names=val_ds.class_indices.keys())\n",
        "print(report)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "1G4s3W_H4HX0",
        "outputId": "c0eaffb0-6632-402f-eff7-5984b645c415"
      },
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "                     precision    recall  f1-score   support\n",
            "\n",
            "Tomato_Early_blight       0.87      0.76      0.81       200\n",
            " Tomato_Late_blight       0.89      0.93      0.91       381\n",
            "     Tomato_healthy       0.97      0.99      0.98       318\n",
            "\n",
            "           accuracy                           0.91       899\n",
            "          macro avg       0.91      0.89      0.90       899\n",
            "       weighted avg       0.91      0.91      0.91       899\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "As expected since, they have similar accuracy scores we see they have similar classification rates. Lets just check a third model, in this case lets look at our most over fit model, model 9. Which has a training accuracy of 100%, but a validation accuracy of 90%."
      ],
      "metadata": {
        "id": "afgVFwVI2WFi"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "y_true_classes, y_pred_classes = model_preds(model9)\n",
        "cm = confusion_matrix(y_true_classes, y_pred_classes)\n",
        "\n",
        "fig, ax = plt.subplots(figsize=(12, 5))\n",
        "\n",
        "sns.heatmap(cm, annot=True, fmt='d')\n",
        "\n",
        "ax.set_yticklabels(['Early Blight', 'Late Blight', 'Healthy'], rotation=0, va=\"center\")\n",
        "ax.set_xticklabels(['Early Blight', 'Late Blight', 'Healthy'], rotation=0, ha=\"center\")\n",
        "plt.title('model 2 matrix')\n",
        "plt.xlabel('Predicted')\n",
        "plt.ylabel('Truth')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 504
        },
        "id": "kl4KTu4C23oK",
        "outputId": "009079cc-d364-421a-fba6-47421f475180"
      },
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "29/29 [==============================] - 3s 88ms/step\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1200x500 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "report = classification_report(y_true_classes, y_pred_classes, target_names=val_ds.class_indices.keys())\n",
        "print(report)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "CLZRNs6J4PPL",
        "outputId": "4ca2ff9d-e00d-4a80-fddc-b033cf7ecc68"
      },
      "execution_count": 27,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "                     precision    recall  f1-score   support\n",
            "\n",
            "Tomato_Early_blight       0.89      0.76      0.82       200\n",
            " Tomato_Late_blight       0.88      0.95      0.92       381\n",
            "     Tomato_healthy       0.99      0.99      0.99       318\n",
            "\n",
            "           accuracy                           0.92       899\n",
            "          macro avg       0.92      0.90      0.91       899\n",
            "       weighted avg       0.92      0.92      0.92       899\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Conclusion\n",
        "\n",
        "Model 8 does pretty well on the accuracy score, but misclassifies a few blight plants as healthy which is something we dont want. Model 8 will beused due to being less overfit since the increase in validation accuracy is marginal. Because it has a low recall score on Early blight, that means we have a risk of missing the early stages of prevenative blight. Missclassifying a healthy plant as unhealthy is not as much of an issue since it would simply require a framer to check up on the plant.\n",
        "\n",
        "The next goal is to improve the recall score of Early blight so farmers can intervenue before the disease spreads and effects other crops."
      ],
      "metadata": {
        "id": "9XWtCRnrzLZy"
      }
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "provenance": [],
      "gpuType": "T4"
    },
    "gpuClass": "standard",
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}