Upload trained model

.gitattributes

@@ -33,3 +33,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+base_model_best.keras filter=lfs diff=lfs merge=lfs -text
+base_model_trained.hdf5 filter=lfs diff=lfs merge=lfs -text
+fine_tune_model_best.keras filter=lfs diff=lfs merge=lfs -text
+fine_tune_model_trained.hdf5 filter=lfs diff=lfs merge=lfs -text

base_model_best.keras

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd9506031065ae9e9721e586262640ac9d57d7262f2866d27072519c01ba198c
+size 243791847

base_model_trained.hdf5

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:21cf4e2830672bbcd62963b1b8b58491b66b2098aebc68e09a9549166e6e6137
+size 243607320

fine_tune_model_best.keras

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:77af2dd4f3918c1c8f327a09aa13b7687b366c0952fc69c049371a7f833d2a7e
+size 308164497

fine_tune_model_trained.hdf5

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d6886051c2f9782bbff94e565556e2323d79810e5dce2076902d31f8038f85d4
+size 307983704

models.ipynb

@@ -0,0 +1,606 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Retrieving speedtest.net configuration...\n",
+      "Testing from VNPT (14.177.252.75)...\n",
+      "Retrieving speedtest.net server list...\n",
+      "Selecting best server based on ping...\n",
+      "Hosted by Viettel IDC (Vinh) [261.88 km]: 10.311 ms\n",
+      "Testing download speed................................................................................\n",
+      "Download: 184.35 Mbit/s\n",
+      "Testing upload speed......................................................................................................\n",
+      "Upload: 202.71 Mbit/s\n"
+     ]
+    }
+   ],
+   "source": [
+    "import tensorflow as tf\n",
+    "import matplotlib.pyplot as plt\n",
+    "import h5py\n",
+    "\n",
+    "tf.get_logger().setLevel('ERROR')\n",
+    "!curl -s https://raw.githubusercontent.com/sivel/speedtest-cli/master/speedtest.py | python -"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "TRAIN_PATH = 'D:/UETCodeCamp/dataset/dataset/Images/Train'\n",
+    "VALIDATE_PATH = 'D:/UETCodeCamp/dataset/dataset/Images/Validate'\n",
+    "TEST_PATH = 'D:/UETCodeCamp/dataset/dataset/Images/Test'"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import os\n",
+    "PATH = 'Models/ResNet152V2'\n",
+    "\n",
+    "BASE_MODEL_BEST = os.path.join(PATH, 'base_model_best.hdf5')\n",
+    "BASE_MODEL_TRAINED = os.path.join(PATH, 'base_model_trained.hdf5')\n",
+    "BASE_MODEL_FIG = os.path.join(PATH, 'base_model_fig.jpg')\n",
+    "\n",
+    "FINE_TUNE_MODEL_BEST = os.path.join(PATH, 'fine_tune_model_best.hdf5')\n",
+    "FINE_TUNE_MODEL_TRAINED = os.path.join(PATH, 'fine_tune_model_trained.hdf5')\n",
+    "FINE_TUNE_MODE_FIG = os.path.join(PATH, 'fine_tune_model_fig.jpg')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "IMAGE_SIZE = (300, 300)\n",
+    "BATCH_SIZE = 128"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
+    "train_generator = ImageDataGenerator(\n",
+    "    rescale = 1./255,\n",
+    "    rotation_range = 40, \n",
+    "    width_shift_range = 0.2, \n",
+    "    height_shift_range = 0.2,\n",
+    "    shear_range = 0.2,\n",
+    "    zoom_range = 0.2,\n",
+    "    horizontal_flip = True\n",
+    ")\n",
+    "validate_generator = ImageDataGenerator(rescale=1./255)\n",
+    "test_generator = ImageDataGenerator(rescale=1./255)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "d:\\UETCodeCamp\\Model\n"
+     ]
+    }
+   ],
+   "source": [
+    "print(os.getcwd())"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Found 18751 images belonging to 38 classes.\n",
+      "Found 2757 images belonging to 38 classes.\n",
+      "Found 5169 images belonging to 38 classes.\n"
+     ]
+    }
+   ],
+   "source": [
+    "generated_train_data = train_generator.flow_from_directory(TRAIN_PATH, target_size=IMAGE_SIZE, batch_size=BATCH_SIZE)\n",
+    "generated_validate_data = validate_generator.flow_from_directory(VALIDATE_PATH, target_size=IMAGE_SIZE, batch_size=BATCH_SIZE)\n",
+    "generated_test_data = test_generator.flow_from_directory(TEST_PATH, target_size=IMAGE_SIZE)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "CLASSES = 38\n",
+    "INITIAL_EPOCHS = 15\n",
+    "FINE_TUNE_EPOCHS = 15\n",
+    "TOTAL_EPOCHS = INITIAL_EPOCHS + FINE_TUNE_EPOCHS\n",
+    "FINE_TUNE_AT = 516"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from tensorflow.keras.applications.resnet_v2 import ResNet152V2\n",
+    "from tensorflow.keras.layers import GlobalAveragePooling2D, Dense, Dropout\n",
+    "from tensorflow.keras.models import Model"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 10,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "pretrained_model = ResNet152V2(weights='imagenet', include_top=False)\n",
+    "last_output = pretrained_model.output\n",
+    "x = GlobalAveragePooling2D()(last_output)\n",
+    "x = Dense(512, activation='relu')(x)\n",
+    "x = Dropout(0.2)(x)\n",
+    "outputs = Dense(CLASSES, activation='softmax')(x)\n",
+    "model = Model(inputs=pretrained_model.input, outputs=outputs)\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 11,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from tensorflow.keras.callbacks import EarlyStopping, ModelCheckpoint\n",
+    "base_checkpointer = ModelCheckpoint(\n",
+    "    filepath=BASE_MODEL_BEST.replace('.hdf5', '.keras'), \n",
+    "    save_best_only=True, \n",
+    "    verbose=1\n",
+    ")\n",
+    "\n",
+    "fine_tune_checkpointer = ModelCheckpoint(\n",
+    "    filepath=FINE_TUNE_MODEL_BEST.replace('.hdf5', '.keras'), \n",
+    "    save_best_only=True,\n",
+    "    verbose=1, \n",
+    ")\n",
+    "\n",
+    "\n",
+    "# Stop if no improvement after 3 epochs\n",
+    "early_stopping = EarlyStopping(monitor='val_loss', patience=3, verbose=1)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 12,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import os\n",
+    "\n",
+    "os.makedirs('Models/ResNet152V2', exist_ok=True)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 13,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "for layer in pretrained_model.layers: layer.trainable = False\n",
+    "model.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 14,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Epoch 1/15\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "c:\\Users\\VuongQuan14\\AppData\\Local\\Programs\\Python\\Python310\\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[1m 55/146\u001b[0m \u001b[32m━━━━━━━\u001b[0m\u001b[37m━━━━━━━━━━━━━\u001b[0m \u001b[1m33:44\u001b[0m 22s/step - accuracy: 0.2474 - loss: 2.9693"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "c:\\Users\\VuongQuan14\\AppData\\Local\\Programs\\Python\\Python310\\lib\\site-packages\\PIL\\Image.py:1056: UserWarning: Palette images with Transparency expressed in bytes should be converted to RGBA images\n",
+      "  warnings.warn(\n"
+     ]
+    },
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "\u001b[1m146/146\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 24s/step - accuracy: 0.3447 - loss: 2.4622 \n",
+      "Epoch 1: val_loss improved from inf to 1.41906, saving model to Models/ResNet152V2\\base_model_best.keras\n",
+      "\u001b[1m146/146\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3788s\u001b[0m 26s/step - accuracy: 0.3453 - loss: 2.4589 - val_accuracy: 0.5796 - val_loss: 1.4191\n",
+      "Epoch 2/15\n",
+      "\u001b[1m  1/146\u001b[0m \u001b[37m━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[1m38:06\u001b[0m 16s/step - accuracy: 0.5469 - loss: 1.6239"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "c:\\Users\\VuongQuan14\\AppData\\Local\\Programs\\Python\\Python310\\lib\\contextlib.py:153: 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": [
+      "\n",
+      "Epoch 2: val_loss did not improve from 1.41906\n",
+      "\u001b[1m146/146\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m24s\u001b[0m 59ms/step - accuracy: 0.5469 - loss: 1.6239 - val_accuracy: 0.5217 - val_loss: 1.6660\n",
+      "Epoch 3/15\n",
+      "\u001b[1m146/146\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 21s/step - accuracy: 0.5845 - loss: 1.3994 \n",
+      "Epoch 3: val_loss improved from 1.41906 to 1.29527, saving model to Models/ResNet152V2\\base_model_best.keras\n",
+      "\u001b[1m146/146\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3664s\u001b[0m 25s/step - accuracy: 0.5846 - loss: 1.3993 - val_accuracy: 0.6131 - val_loss: 1.2953\n",
+      "Epoch 4/15\n",
+      "\u001b[1m  1/146\u001b[0m \u001b[37m━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[1m1:11:06\u001b[0m 29s/step - accuracy: 0.5938 - loss: 1.2649\n",
+      "Epoch 4: val_loss improved from 1.29527 to 1.22025, saving model to Models/ResNet152V2\\base_model_best.keras\n",
+      "\u001b[1m146/146\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m50s\u001b[0m 143ms/step - accuracy: 0.5938 - loss: 1.2649 - val_accuracy: 0.6667 - val_loss: 1.2203\n",
+      "Epoch 5/15\n",
+      "\u001b[1m146/146\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 21s/step - accuracy: 0.6200 - loss: 1.2513 \n",
+      "Epoch 5: val_loss improved from 1.22025 to 1.17931, saving model to Models/ResNet152V2\\base_model_best.keras\n",
+      "\u001b[1m146/146\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3524s\u001b[0m 24s/step - accuracy: 0.6201 - loss: 1.2513 - val_accuracy: 0.6481 - val_loss: 1.1793\n",
+      "Epoch 6/15\n",
+      "\u001b[1m  1/146\u001b[0m \u001b[37m━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[1m38:32\u001b[0m 16s/step - accuracy: 0.6484 - loss: 1.1963\n",
+      "Epoch 6: val_loss improved from 1.17931 to 1.05118, saving model to Models/ResNet152V2\\base_model_best.keras\n",
+      "\u001b[1m146/146\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m30s\u001b[0m 95ms/step - accuracy: 0.6484 - loss: 1.1963 - val_accuracy: 0.6232 - val_loss: 1.0512\n",
+      "Epoch 7/15\n",
+      "\u001b[1m146/146\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 44s/step - accuracy: 0.6569 - loss: 1.1265 \n",
+      "Epoch 7: val_loss did not improve from 1.05118\n",
+      "\u001b[1m146/146\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8147s\u001b[0m 56s/step - accuracy: 0.6569 - loss: 1.1265 - val_accuracy: 0.6715 - val_loss: 1.1207\n",
+      "Epoch 8/15\n",
+      "\u001b[1m  1/146\u001b[0m \u001b[37m━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[1m4:56:50\u001b[0m 123s/step - accuracy: 0.7188 - loss: 0.9636\n",
+      "Epoch 8: val_loss did not improve from 1.05118\n",
+      "\u001b[1m146/146\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m192s\u001b[0m 475ms/step - accuracy: 0.7188 - loss: 0.9636 - val_accuracy: 0.6667 - val_loss: 1.1370\n",
+      "Epoch 9/15\n",
+      "\u001b[1m146/146\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 20s/step - accuracy: 0.6758 - loss: 1.0515 \n",
+      "Epoch 9: val_loss did not improve from 1.05118\n",
+      "\u001b[1m146/146\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3502s\u001b[0m 23s/step - accuracy: 0.6758 - loss: 1.0515 - val_accuracy: 0.6864 - val_loss: 1.0649\n",
+      "Epoch 9: early stopping\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "WARNING:absl:You are saving your model as an HDF5 file via `model.save()` or `keras.saving.save_model(model)`. This file format is considered legacy. We recommend using instead the native Keras format, e.g. `model.save('my_model.keras')` or `keras.saving.save_model(model, 'my_model.keras')`. \n"
+     ]
+    }
+   ],
+   "source": [
+    "history = model.fit(\n",
+    "    generated_train_data,\n",
+    "    validation_data = generated_validate_data,\n",
+    "    validation_steps = generated_validate_data.n // BATCH_SIZE,\n",
+    "    steps_per_epoch = generated_train_data.n // BATCH_SIZE,\n",
+    "    callbacks = [base_checkpointer, early_stopping],\n",
+    "    epochs = INITIAL_EPOCHS,\n",
+    "    verbose = 1,\n",
+    ")\n",
+    "model.save(BASE_MODEL_TRAINED)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 15,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "(300, 300, 3)\n",
+      "{'Banh beo': 0, 'Banh bot loc': 1, 'Banh can': 2, 'Banh canh': 3, 'Banh chung': 4, 'Banh cuon': 5, 'Banh duc': 6, 'Banh gio': 7, 'Banh khot': 8, 'Banh mi': 9, 'Banh pia': 10, 'Banh tet': 11, 'Banh trang nuong': 12, 'Banh xeo': 13, 'Bun bo Hue': 14, 'Bun dau mam tom': 15, 'Bun mam': 16, 'Bun rieu': 17, 'Bun thit nuong': 18, 'Bánh cu đơ': 19, 'Bánh mì cay': 20, 'Bánh đa cua': 21, 'Bánh đậu xanh': 22, 'Bò bía': 23, 'Bún cá': 24, 'Ca kho to': 25, 'Canh chua': 26, 'Cao lau': 27, 'Chao long': 28, 'Com tam': 29, 'Cơm cháy': 30, 'Goi cuon': 31, 'Hu tieu': 32, 'Mi quang': 33, 'Nem chua': 34, 'Nem nướng': 35, 'Pho': 36, 'Xoi xeo': 37}\n"
+     ]
+    }
+   ],
+   "source": [
+    "print(generated_train_data.image_shape)\n",
+    "print(generated_train_data.class_indices)\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 16,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "\n",
+    "for layer in pretrained_model.layers[:FINE_TUNE_AT]: layer.trainable = False\n",
+    "for layer in pretrained_model.layers[FINE_TUNE_AT:]: layer.trainable = True"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 17,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from tensorflow.keras.optimizers import SGD\n",
+    "model.compile(\n",
+    "    optimizer = SGD(learning_rate=1e-4, momentum=0.9), \n",
+    "    loss = 'categorical_crossentropy', \n",
+    "    metrics = ['accuracy']\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 18,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Epoch 9/30\n",
+      "\u001b[1m146/146\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 25s/step - accuracy: 0.5182 - loss: 1.7415 \n",
+      "Epoch 9: val_loss improved from inf to 1.11962, saving model to Models/ResNet152V2\\fine_tune_model_best.keras\n",
+      "\u001b[1m146/146\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4345s\u001b[0m 30s/step - accuracy: 0.5186 - loss: 1.7399 - val_accuracy: 0.6819 - val_loss: 1.1196\n",
+      "Epoch 10/30\n",
+      "\u001b[1m  1/146\u001b[0m \u001b[37m━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[1m1:22:53\u001b[0m 34s/step - accuracy: 0.7031 - loss: 1.0933\n",
+      "Epoch 10: val_loss did not improve from 1.11962\n",
+      "\u001b[1m146/146\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m51s\u001b[0m 112ms/step - accuracy: 0.7031 - loss: 1.0933 - val_accuracy: 0.6667 - val_loss: 1.5291\n",
+      "Epoch 11/30\n",
+      "\u001b[1m146/146\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 33s/step - accuracy: 0.6629 - loss: 1.1745 \n",
+      "Epoch 11: val_loss improved from 1.11962 to 1.06462, saving model to Models/ResNet152V2\\fine_tune_model_best.keras\n",
+      "\u001b[1m146/146\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5206s\u001b[0m 35s/step - accuracy: 0.6630 - loss: 1.1742 - val_accuracy: 0.6886 - val_loss: 1.0646\n",
+      "Epoch 11: early stopping\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "WARNING:absl:You are saving your model as an HDF5 file via `model.save()` or `keras.saving.save_model(model)`. This file format is considered legacy. We recommend using instead the native Keras format, e.g. `model.save('my_model.keras')` or `keras.saving.save_model(model, 'my_model.keras')`. \n"
+     ]
+    }
+   ],
+   "source": [
+    "history_fine = model.fit(\n",
+    "    generated_train_data,\n",
+    "    validation_data = generated_validate_data,\n",
+    "    validation_steps = generated_validate_data.n // BATCH_SIZE,\n",
+    "    steps_per_epoch = generated_train_data.n // BATCH_SIZE,\n",
+    "    epochs = TOTAL_EPOCHS,\n",
+    "    initial_epoch = history.epoch[-1],\n",
+    "    callbacks = [fine_tune_checkpointer, early_stopping],\n",
+    "    verbose = 1,\n",
+    ")\n",
+    "model.save(FINE_TUNE_MODEL_TRAINED)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 19,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "\u001b[1m162/162\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1063s\u001b[0m 7s/step - accuracy: 0.6897 - loss: 1.0454\n",
+      "Test accuracy: 0.6889146566390991\n"
+     ]
+    }
+   ],
+   "source": [
+    "loss, accuracy = model.evaluate(generated_test_data)\n",
+    "print('Test accuracy:', accuracy)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 22,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Requirement already satisfied: scipy in c:\\users\\vuongquan14\\appdata\\roaming\\python\\python310\\site-packages (1.10.1)\n",
+      "Collecting scipy\n",
+      "  Downloading scipy-1.14.0-cp310-cp310-win_amd64.whl.metadata (60 kB)\n",
+      "Requirement already satisfied: numpy<2.3,>=1.23.5 in c:\\users\\vuongquan14\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from scipy) (1.23.5)\n",
+      "Downloading scipy-1.14.0-cp310-cp310-win_amd64.whl (44.8 MB)\n",
+      "   ---------------------------------------- 0.0/44.8 MB ? eta -:--:--\n",
+      "    --------------------------------------- 1.0/44.8 MB 7.1 MB/s eta 0:00:07\n",
+      "   ---- ----------------------------------- 5.0/44.8 MB 13.7 MB/s eta 0:00:03\n",
+      "   ----------- ---------------------------- 12.6/44.8 MB 22.5 MB/s eta 0:00:02\n",
+      "   ------------------ --------------------- 20.2/44.8 MB 26.0 MB/s eta 0:00:01\n",
+      "   ------------------------ --------------- 27.0/44.8 MB 27.1 MB/s eta 0:00:01\n",
+      "   ------------------------------ --------- 34.1/44.8 MB 28.1 MB/s eta 0:00:01\n",
+      "   ------------------------------------- -- 41.7/44.8 MB 29.5 MB/s eta 0:00:01\n",
+      "   ---------------------------------------- 44.8/44.8 MB 27.9 MB/s eta 0:00:00\n",
+      "Installing collected packages: scipy\n",
+      "  Attempting uninstall: scipy\n",
+      "    Found existing installation: scipy 1.10.1\n",
+      "    Uninstalling scipy-1.10.1:\n",
+      "      Successfully uninstalled scipy-1.10.1\n",
+      "  Rolling back uninstall of scipy\n",
+      "  Moving to c:\\users\\vuongquan14\\appdata\\roaming\\python\\python310\\site-packages\\scipy-1.10.1-cp310-cp310-win_amd64.whl\n",
+      "   from C:\\Users\\VuongQuan14\\AppData\\Local\\Temp\\pip-uninstall-bxl8tsg7\\scipy-1.10.1-cp310-cp310-win_amd64.whl\n",
+      "  Moving to c:\\users\\vuongquan14\\appdata\\roaming\\python\\python310\\site-packages\\scipy-1.10.1.dist-info\\\n",
+      "   from C:\\Users\\VuongQuan14\\AppData\\Roaming\\Python\\Python310\\site-packages\\~cipy-1.10.1.dist-info\n",
+      "  Moving to c:\\users\\vuongquan14\\appdata\\roaming\\python\\python310\\site-packages\\scipy.libs\\\n",
+      "   from C:\\Users\\VuongQuan14\\AppData\\Roaming\\Python\\Python310\\site-packages\\~cipy.libs\n",
+      "  Moving to c:\\users\\vuongquan14\\appdata\\roaming\\python\\python310\\site-packages\\scipy\\\n",
+      "   from C:\\Users\\VuongQuan14\\AppData\\Roaming\\Python\\Python310\\site-packages\\~cipy\n",
+      "Note: you may need to restart the kernel to use updated packages.\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "WARNING: Ignoring invalid distribution -cipy (c:\\users\\vuongquan14\\appdata\\local\\programs\\python\\python310\\lib\\site-packages)\n",
+      "WARNING: Ignoring invalid distribution -cipy (c:\\users\\vuongquan14\\appdata\\local\\programs\\python\\python310\\lib\\site-packages)\n",
+      "ERROR: Could not install packages due to an OSError: [WinError 5] Access is denied: 'c:\\\\Users\\\\VuongQuan14\\\\AppData\\\\Local\\\\Programs\\\\Python\\\\Python310\\\\Lib\\\\site-packages\\\\scipy\\\\linalg\\\\cython_blas.cp310-win_amd64.pyd'\n",
+      "Consider using the `--user` option or check the permissions.\n",
+      "\n"
+     ]
+    }
+   ],
+   "source": [
+    "pip install --upgrade scipy\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 23,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Found 5169 images belonging to 38 classes.\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "c:\\Users\\VuongQuan14\\AppData\\Local\\Programs\\Python\\Python310\\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[1m41/41\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m883s\u001b[0m 22s/step\n",
+      "                  precision    recall  f1-score   support\n",
+      "\n",
+      "        Banh beo       0.84      0.71      0.77       129\n",
+      "    Banh bot loc       0.60      0.61      0.60       144\n",
+      "        Banh can       0.83      0.69      0.75       149\n",
+      "       Banh canh       0.44      0.37      0.40       193\n",
+      "      Banh chung       0.80      0.77      0.79       102\n",
+      "       Banh cuon       0.70      0.68      0.69       228\n",
+      "        Banh duc       0.41      0.20      0.27       133\n",
+      "        Banh gio       0.75      0.81      0.78       129\n",
+      "       Banh khot       0.69      0.83      0.76       167\n",
+      "         Banh mi       0.92      0.91      0.91       268\n",
+      "        Banh pia       0.86      0.84      0.85        89\n",
+      "        Banh tet       0.83      0.73      0.78       138\n",
+      "Banh trang nuong       0.90      0.75      0.82       159\n",
+      "        Banh xeo       0.81      0.83      0.82       235\n",
+      "      Bun bo Hue       0.54      0.68      0.60       306\n",
+      " Bun dau mam tom       0.90      0.90      0.90       184\n",
+      "         Bun mam       0.62      0.61      0.62       155\n",
+      "        Bun rieu       0.54      0.68      0.60       231\n",
+      "  Bun thit nuong       0.57      0.65      0.61       150\n",
+      "      Bánh cu đơ       0.68      0.72      0.70        18\n",
+      "     Bánh mì cay       0.82      0.64      0.72        14\n",
+      "     Bánh đa cua       0.50      0.21      0.30        14\n",
+      "   Bánh đậu xanh       0.91      0.59      0.71        17\n",
+      "          Bò bía       0.57      0.42      0.48        19\n",
+      "          Bún cá       0.00      0.00      0.00        13\n",
+      "       Ca kho to       0.86      0.86      0.86       136\n",
+      "       Canh chua       0.62      0.63      0.62       165\n",
+      "         Cao lau       0.60      0.67      0.63       124\n",
+      "       Chao long       0.71      0.73      0.72       215\n",
+      "         Com tam       0.77      0.81      0.79       189\n",
+      "        Cơm cháy       0.81      0.72      0.76        18\n",
+      "        Goi cuon       0.78      0.78      0.78       172\n",
+      "         Hu tieu       0.46      0.38      0.42       197\n",
+      "        Mi quang       0.55      0.75      0.63       177\n",
+      "        Nem chua       0.67      0.55      0.61       109\n",
+      "       Nem nướng       0.56      0.56      0.56        16\n",
+      "             Pho       0.55      0.48      0.51       162\n",
+      "         Xoi xeo       0.89      0.77      0.83       105\n",
+      "\n",
+      "        accuracy                           0.69      5169\n",
+      "       macro avg       0.68      0.65      0.66      5169\n",
+      "    weighted avg       0.69      0.69      0.69      5169\n",
+      "\n"
+     ]
+    }
+   ],
+   "source": [
+    "from sklearn.metrics import classification_report\n",
+    "import numpy as np\n",
+    "\n",
+    "\n",
+    "# Sử dụng generator để dự đoán nhãn cho dữ liệu kiểm tra\n",
+    "generated_test_data = test_generator.flow_from_directory(TEST_PATH, target_size=IMAGE_SIZE, batch_size=BATCH_SIZE, shuffle=False)\n",
+    "\n",
+    "# Dự đoán nhãn\n",
+    "predictions = model.predict(generated_test_data)\n",
+    "y_pred = np.argmax(predictions, axis=1)\n",
+    "y_true = generated_test_data.classes\n",
+    "\n",
+    "# Tính toán và in ra các chỉ số\n",
+    "class_labels = list(generated_test_data.class_indices.keys())\n",
+    "report = classification_report(y_true, y_pred, target_names=class_labels)\n",
+    "print(report)\n"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "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": "3.10.9"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}