jan24_cds_mushrooms/notebooks/transfer_learning_ResNet50/ResNet50_L2.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 2199 images belonging to 2 classes.\n",
      "Found 549 images belonging to 2 classes.\n",
      "Training samples: 2199\n",
      "Validation samples: 549\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
    "\n",
    "dataset_dir = 'C:/Users/vsavelev/GITHUB/DS_projet/jan24_cds_mushrooms/data'\n",
    "\n",
    "# Create ImageDataGenerator with validation split\n",
    "datagen = ImageDataGenerator(rescale=1.0/255, validation_split=0.2)\n",
    "\n",
    "train_generator = datagen.flow_from_directory(\n",
    "    dataset_dir,\n",
    "    target_size=(224, 224),\n",
    "    batch_size=32,\n",
    "    class_mode='categorical',\n",
    "    subset='training'  # Set as training data\n",
    ")\n",
    "\n",
    "validation_generator = datagen.flow_from_directory(\n",
    "    dataset_dir,\n",
    "    target_size=(224, 224),\n",
    "    batch_size=32,\n",
    "    class_mode='categorical',\n",
    "    subset='validation'  # Set as validation data\n",
    ")\n",
    "\n",
    "print(f'Training samples: {train_generator.samples}')\n",
    "print(f'Validation samples: {validation_generator.samples}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\">Model: \"sequential\"</span>\n",
       "</pre>\n"
      ],
      "text/plain": [
       "\u001b[1mModel: \"sequential\"\u001b[0m\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\">┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n",
       "┃<span style=\"font-weight: bold\"> Layer (type)                    </span>┃<span style=\"font-weight: bold\"> Output Shape           </span>┃<span style=\"font-weight: bold\">       Param # </span>┃\n",
       "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n",
       "│ resnet50 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Functional</span>)           │ ?                      │    <span style=\"color: #00af00; text-decoration-color: #00af00\">23,587,712</span> │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ global_average_pooling2d        │ ?                      │   <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (unbuilt) │\n",
       "│ (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">GlobalAveragePooling2D</span>)        │                        │               │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dense (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dense</span>)                   │ ?                      │   <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (unbuilt) │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dropout (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dropout</span>)               │ ?                      │   <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (unbuilt) │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dense_1 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dense</span>)                 │ ?                      │   <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (unbuilt) │\n",
       "└─────────────────────────────────┴────────────────────────┴───────────────┘\n",
       "</pre>\n"
      ],
      "text/plain": [
       "┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n",
       "┃\u001b[1m \u001b[0m\u001b[1mLayer (type)                   \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1mOutput Shape          \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1m      Param #\u001b[0m\u001b[1m \u001b[0m┃\n",
       "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n",
       "│ resnet50 (\u001b[38;5;33mFunctional\u001b[0m)           │ ?                      │    \u001b[38;5;34m23,587,712\u001b[0m │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ global_average_pooling2d        │ ?                      │   \u001b[38;5;34m0\u001b[0m (unbuilt) │\n",
       "│ (\u001b[38;5;33mGlobalAveragePooling2D\u001b[0m)        │                        │               │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dense (\u001b[38;5;33mDense\u001b[0m)                   │ ?                      │   \u001b[38;5;34m0\u001b[0m (unbuilt) │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dropout (\u001b[38;5;33mDropout\u001b[0m)               │ ?                      │   \u001b[38;5;34m0\u001b[0m (unbuilt) │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dense_1 (\u001b[38;5;33mDense\u001b[0m)                 │ ?                      │   \u001b[38;5;34m0\u001b[0m (unbuilt) │\n",
       "└─────────────────────────────────┴────────────────────────┴───────────────┘\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Total params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">23,587,712</span> (89.98 MB)\n",
       "</pre>\n"
      ],
      "text/plain": [
       "\u001b[1m Total params: \u001b[0m\u001b[38;5;34m23,587,712\u001b[0m (89.98 MB)\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Trainable params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (0.00 B)\n",
       "</pre>\n"
      ],
      "text/plain": [
       "\u001b[1m Trainable params: \u001b[0m\u001b[38;5;34m0\u001b[0m (0.00 B)\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Non-trainable params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">23,587,712</span> (89.98 MB)\n",
       "</pre>\n"
      ],
      "text/plain": [
       "\u001b[1m Non-trainable params: \u001b[0m\u001b[38;5;34m23,587,712\u001b[0m (89.98 MB)\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from tensorflow.keras.applications import ResNet50\n",
    "from tensorflow.keras import layers, models\n",
    "\n",
    "# Load and Configure the Pre-trained ResNet50 Model\n",
    "base_model = ResNet50(weights='imagenet', include_top=False, input_shape=(224, 224, 3))\n",
    "\"\"\"\n",
    "weights='imagenet': Loads the pre-trained weights from the ImageNet dataset.\n",
    "include_top=False: Excludes the top fully-connected layers of the ResNet50 model, enabling you to add your own custom layers.\n",
    "input_shape=(224, 224, 3): Specifies the input shape of the images (224x224 pixels, with 3 color channels - RGB).\n",
    "\"\"\"\n",
    "\n",
    "# Freeze the base model (to freeze the pre-trained layers)\n",
    "base_model.trainable = False\n",
    "\n",
    "# Add custom layers on top of the base model\n",
    "model = models.Sequential([   #allows to stack layers linearly\n",
    "    base_model,\n",
    "    layers.GlobalAveragePooling2D(),\n",
    "    layers.Dense(1024, activation='relu', kernel_regularizer=tf.keras.regularizers.l2(0.01)),  # L2 regularization\n",
    "    layers.Dropout(0.5),\n",
    "    layers.Dense(train_generator.num_classes, activation='softmax')\n",
    "])\n",
    "\n",
    "\"\"\"\n",
    "GlobalAveragePooling2D(): Reduces each feature map to a single number by taking the average, \n",
    "which helps to reduce the size of the model and prevent overfitting.\n",
    "Dense(1024, activation='relu'): Adds a fully connected layer with 1024 units and ReLU activation function.\n",
    "Dropout(0.5): Adds a dropout layer with a 50% dropout rate to prevent overfitting by randomly setting half of the input units \n",
    "to 0 at each update during training.\n",
    "Dense(train_generator.num_classes, activation='softmax'): \n",
    "Adds the final output layer with units equal to the number of classes in your dataset, using the softmax activation function for multi-class classification.\n",
    "\"\"\"\n",
    "model.compile(optimizer=tf.keras.optimizers.Adam(),\n",
    "              loss='categorical_crossentropy',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "\"\"\"\n",
    "optimizer=tf.keras.optimizers.Adam(): Uses the Adam optimizer, which is an adaptive learning rate optimization algorithm.\n",
    "loss='categorical_crossentropy': Uses categorical cross-entropy as the loss function, suitable for multi-class classification.\n",
    "metrics=['accuracy']: Tracks accuracy as the metric to evaluate the model's performance during training and testing.\n",
    "\"\"\"\n",
    "model.summary()\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\Users\\vsavelev\\AppData\\Local\\anaconda3\\Lib\\site-packages\\keras\\src\\trainers\\data_adapters\\py_dataset_adapter.py:121: UserWarning: Your `PyDataset` class should call `super().__init__(**kwargs)` in its constructor. `**kwargs` can include `workers`, `use_multiprocessing`, `max_queue_size`. Do not pass these arguments to `fit()`, as they will be ignored.\n",
      "  self._warn_if_super_not_called()\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m143s\u001b[0m 2s/step - accuracy: 0.8540 - loss: 6.8043 - val_accuracy: 0.8860 - val_loss: 0.6636\n",
      "Epoch 2/10\n",
      "\u001b[1m 1/68\u001b[0m \u001b[37m━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[1m2:09\u001b[0m 2s/step - accuracy: 0.8438 - loss: 0.8550"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\Users\\vsavelev\\AppData\\Local\\anaconda3\\Lib\\contextlib.py:158: UserWarning: Your input ran out of data; interrupting training. Make sure that your dataset or generator can generate at least `steps_per_epoch * epochs` batches. You may need to use the `.repeat()` function when building your dataset.\n",
      "  self.gen.throw(typ, value, traceback)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 11ms/step - accuracy: 0.8438 - loss: 0.8550 - val_accuracy: 1.0000 - val_loss: 0.4526\n",
      "Epoch 3/10\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m146s\u001b[0m 2s/step - accuracy: 0.8873 - loss: 0.6507 - val_accuracy: 0.8879 - val_loss: 0.4938\n",
      "Epoch 4/10\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 6ms/step - accuracy: 0.8750 - loss: 0.5112 - val_accuracy: 0.8000 - val_loss: 0.6924\n",
      "Epoch 5/10\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m149s\u001b[0m 2s/step - accuracy: 0.8857 - loss: 0.5040 - val_accuracy: 0.8879 - val_loss: 0.4471\n",
      "Epoch 6/10\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 6ms/step - accuracy: 0.8750 - loss: 0.4322 - val_accuracy: 0.8000 - val_loss: 0.6413\n",
      "Epoch 7/10\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m166s\u001b[0m 2s/step - accuracy: 0.8926 - loss: 0.4378 - val_accuracy: 0.8860 - val_loss: 0.4396\n",
      "Epoch 8/10\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.9375 - loss: 0.3477 - val_accuracy: 1.0000 - val_loss: 0.2697\n",
      "Epoch 9/10\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m173s\u001b[0m 3s/step - accuracy: 0.8890 - loss: 0.4132 - val_accuracy: 0.8879 - val_loss: 0.4444\n",
      "Epoch 10/10\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.9688 - loss: 0.3492 - val_accuracy: 0.8000 - val_loss: 0.5367\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(\n",
    "    train_generator,\n",
    "    steps_per_epoch=train_generator.samples // train_generator.batch_size,\n",
    "    validation_data=validation_generator,\n",
    "    validation_steps=validation_generator.samples // validation_generator.batch_size,\n",
    "    epochs=10\n",
    ")\n",
    "\n",
    "#This specifies the number of complete passes through the training dataset. Here, the model will train for 10 epochs."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model compiled successfully.\n",
      "Callbacks created successfully.\n",
      "Epoch 1/20\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m267s\u001b[0m 4s/step - accuracy: 0.8686 - loss: 0.6860 - val_accuracy: 0.8897 - val_loss: 0.5719\n",
      "Epoch 2/20\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 9ms/step - accuracy: 0.8750 - loss: 0.6368 - val_accuracy: 0.6000 - val_loss: 0.7365\n",
      "Epoch 3/20\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m286s\u001b[0m 4s/step - accuracy: 0.8856 - loss: 0.6046 - val_accuracy: 0.8897 - val_loss: 0.6145\n",
      "Epoch 4/20\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 9ms/step - accuracy: 0.8438 - loss: 0.5484 - val_accuracy: 0.6000 - val_loss: 0.7114\n",
      "Epoch 5/20\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m323s\u001b[0m 5s/step - accuracy: 0.8840 - loss: 0.5082 - val_accuracy: 0.8860 - val_loss: 0.5360\n",
      "Epoch 6/20\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 7ms/step - accuracy: 0.8750 - loss: 0.4655 - val_accuracy: 1.0000 - val_loss: 0.4314\n",
      "Epoch 7/20\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m299s\u001b[0m 4s/step - accuracy: 0.8770 - loss: 0.4487 - val_accuracy: 0.8879 - val_loss: 0.4546\n",
      "Epoch 8/20\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 6ms/step - accuracy: 0.9375 - loss: 0.3386 - val_accuracy: 0.8000 - val_loss: 0.5507\n",
      "Epoch 9/20\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m276s\u001b[0m 4s/step - accuracy: 0.8946 - loss: 0.3906 - val_accuracy: 0.8879 - val_loss: 0.4199\n",
      "Epoch 10/20\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 7ms/step - accuracy: 0.8438 - loss: 0.4413 - val_accuracy: 0.8000 - val_loss: 0.5716\n",
      "Epoch 11/20\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m293s\u001b[0m 4s/step - accuracy: 0.8840 - loss: 0.3867 - val_accuracy: 0.8860 - val_loss: 0.4153\n",
      "Epoch 12/20\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 7ms/step - accuracy: 0.9565 - loss: 0.2604 - val_accuracy: 1.0000 - val_loss: 0.2164\n",
      "Epoch 13/20\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m290s\u001b[0m 4s/step - accuracy: 0.8951 - loss: 0.3485 - val_accuracy: 0.8915 - val_loss: 0.3989\n",
      "Epoch 14/20\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 7ms/step - accuracy: 0.9375 - loss: 0.2921 - val_accuracy: 0.4000 - val_loss: 1.3309\n",
      "Epoch 15/20\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m273s\u001b[0m 4s/step - accuracy: 0.8950 - loss: 0.3413 - val_accuracy: 0.8860 - val_loss: 0.4095\n",
      "Epoch 16/20\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 7ms/step - accuracy: 0.9062 - loss: 0.3136 - val_accuracy: 1.0000 - val_loss: 0.1990\n",
      "Epoch 17/20\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m270s\u001b[0m 4s/step - accuracy: 0.8919 - loss: 0.3295 - val_accuracy: 0.8860 - val_loss: 0.4054\n",
      "Epoch 18/20\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - accuracy: 0.8750 - loss: 0.3683 - val_accuracy: 1.0000 - val_loss: 0.1749\n",
      "Epoch 19/20\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m279s\u001b[0m 4s/step - accuracy: 0.8906 - loss: 0.3170 - val_accuracy: 0.8860 - val_loss: 0.4067\n",
      "Epoch 20/20\n",
      "\u001b[1m68/68\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - accuracy: 0.9375 - loss: 0.2421 - val_accuracy: 1.0000 - val_loss: 0.2140\n"
     ]
    }
   ],
   "source": [
    "from tensorflow.keras.callbacks import EarlyStopping, ModelCheckpoint\n",
    "\n",
    "\n",
    "# Unfreeze some layers\n",
    "base_model.trainable = True\n",
    "fine_tune_at = 100  # fine-tune from this layer onwards\n",
    "\n",
    "for layer in base_model.layers[:fine_tune_at]:\n",
    "    layer.trainable = False\n",
    "\n",
    "#  # Unfreeze more layers gradually\n",
    "# for layer in base_model.layers[:-10]:  # Unfreeze all layers except the last 10 layers\n",
    "#     layer.trainable = False\n",
    "\n",
    "model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=1e-6), #learning_rate=1e-5\n",
    "              loss='categorical_crossentropy', #The categorical cross-entropy loss function is used because this is a multi-class classification problem\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "print(\"Model compiled successfully.\")\n",
    "\n",
    "\"\"\"\n",
    "The Adam optimizer is used with a very small learning rate (1e-5). Fine-tuning typically \n",
    "uses a smaller learning rate to prevent large updates to the weights, which could potentially destroy the learned features in the pre-trained model.\n",
    "\"\"\"\n",
    "\n",
    "early_stopping = EarlyStopping(monitor='val_loss', patience=5, restore_best_weights=True)\n",
    "model_checkpoint = ModelCheckpoint('best_model.keras', save_best_only=True, monitor='val_loss')\n",
    "\n",
    "print(\"Callbacks created successfully.\")\n",
    "\n",
    "\n",
    "history_fine = model.fit(\n",
    "    train_generator,\n",
    "    steps_per_epoch=train_generator.samples // train_generator.batch_size,\n",
    "    validation_data=validation_generator,\n",
    "    validation_steps=validation_generator.samples // validation_generator.batch_size,\n",
    "    epochs=20,\n",
    "    #callbacks=[early_stopping, model_checkpoint]\n",
    ")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1m18/18\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m22s\u001b[0m 1s/step - accuracy: 0.8704 - loss: 0.4312\n",
      "Validation loss: 0.4050602614879608\n",
      "Validation accuracy: 0.8870673775672913\n"
     ]
    }
   ],
   "source": [
    "loss, accuracy = model.evaluate(validation_generator)\n",
    "print(f'Validation loss: {loss}')\n",
    "print(f'Validation accuracy: {accuracy}')\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "base",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}