"
]
},
"execution_count": 26,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"model.fit(\n",
" train_dataset,\n",
" validation_data=validation_dataset,\n",
" epochs=EPOCHS\n",
" )"
]
},
{
"cell_type": "markdown",
"id": "ea12f4e5",
"metadata": {},
"source": [
"## Grok.com's recommendation"
]
},
{
"cell_type": "code",
"execution_count": 27,
"id": "151c4432",
"metadata": {},
"outputs": [],
"source": [
"data_argumentation=tf.keras.Sequential([\n",
" tf.keras.layers.RandomFlip(\"horizontal\"),\n",
" tf.keras.layers.RandomRotation(0.15),\n",
" tf.keras.layers.RandomZoom(0.15)\n",
"])"
]
},
{
"cell_type": "code",
"execution_count": 29,
"id": "752ca3de",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Epoch 1/50\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m67s\u001b[0m 2s/step - accuracy: 0.3037 - loss: 2.9248 - val_accuracy: 0.2429 - val_loss: 3.7958 - learning_rate: 3.0000e-04\n",
"Epoch 2/50\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m58s\u001b[0m 3s/step - accuracy: 0.4167 - loss: 2.5839 - val_accuracy: 0.2429 - val_loss: 5.5037 - learning_rate: 3.0000e-04\n",
"Epoch 3/50\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m57s\u001b[0m 2s/step - accuracy: 0.4124 - loss: 2.6607 - val_accuracy: 0.2429 - val_loss: 4.8246 - learning_rate: 3.0000e-04\n",
"Epoch 4/50\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 2s/step - accuracy: 0.5055 - loss: 2.3793\n",
"Epoch 4: ReduceLROnPlateau reducing learning rate to 0.0001500000071246177.\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m57s\u001b[0m 2s/step - accuracy: 0.4675 - loss: 2.4681 - val_accuracy: 0.2429 - val_loss: 5.5806 - learning_rate: 3.0000e-04\n",
"Epoch 5/50\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m58s\u001b[0m 3s/step - accuracy: 0.4590 - loss: 2.4254 - val_accuracy: 0.2429 - val_loss: 6.4763 - learning_rate: 1.5000e-04\n",
"Epoch 6/50\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m58s\u001b[0m 2s/step - accuracy: 0.5042 - loss: 2.3246 - val_accuracy: 0.2429 - val_loss: 7.2424 - learning_rate: 1.5000e-04\n",
"Epoch 7/50\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 2s/step - accuracy: 0.4722 - loss: 2.4259\n",
"Epoch 7: ReduceLROnPlateau reducing learning rate to 7.500000356230885e-05.\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m58s\u001b[0m 3s/step - accuracy: 0.4986 - loss: 2.3835 - val_accuracy: 0.2429 - val_loss: 5.2781 - learning_rate: 1.5000e-04\n",
"Epoch 8/50\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m61s\u001b[0m 3s/step - accuracy: 0.5113 - loss: 2.3075 - val_accuracy: 0.2429 - val_loss: 5.3866 - learning_rate: 7.5000e-05\n",
"Epoch 8: early stopping\n",
"Restoring model weights from the end of the best epoch: 1.\n",
"\n",
" Training Complete\n",
"Best Validation accuracy :0.2429\n"
]
}
],
"source": [
"model=tf.keras.models.Sequential()\n",
"\n",
"model.add(data_argumentation)\n",
"model.add(tf.keras.layers.Rescaling(1./255,input_shape=(IMG_HEIGHT,IMG_WIDTH,3)))\n",
"\n",
"model.add(tf.keras.layers.Conv2D(32,(3,3),activation='relu',padding='same'))\n",
"model.add(tf.keras.layers.BatchNormalization())\n",
"model.add(tf.keras.layers.Conv2D(32,(3,3),activation='relu',padding='same'))\n",
"model.add(tf.keras.layers.BatchNormalization())\n",
"model.add(tf.keras.layers.MaxPooling2D((2,2)))\n",
"model.add(tf.keras.layers.Dropout(0.25))\n",
"\n",
"\n",
"model.add(tf.keras.layers.Conv2D(64,(3,3),activation='relu',padding='same'))\n",
"model.add(tf.keras.layers.BatchNormalization())\n",
"model.add(tf.keras.layers.Conv2D(64,(3,3),activation='relu',padding='same'))\n",
"model.add(tf.keras.layers.BatchNormalization())\n",
"model.add(tf.keras.layers.MaxPooling2D((2,2)))\n",
"model.add(tf.keras.layers.Dropout(0.25))\n",
"\n",
"\n",
"\n",
"model.add(tf.keras.layers.Conv2D(128,(3,3),activation='relu',padding='same'))\n",
"model.add(tf.keras.layers.BatchNormalization())\n",
"model.add(tf.keras.layers.Conv2D(128,(3,3),activation='relu',padding='same'))\n",
"model.add(tf.keras.layers.BatchNormalization())\n",
"model.add(tf.keras.layers.MaxPooling2D((2,2)))\n",
"model.add(tf.keras.layers.Dropout(0.35))\n",
"\n",
"\n",
"model.add(tf.keras.layers.Flatten())\n",
"\n",
"\n",
"model.add(tf.keras.layers.Dense(256,activation='relu',kernel_regularizer=tf.keras.regularizers.l2(0.001)))\n",
"model.add(tf.keras.layers.BatchNormalization())\n",
"model.add(tf.keras.layers.Dropout(0.5))\n",
"\n",
"\n",
"model.add(tf.keras.layers.Dense(128,activation='relu',kernel_regularizer=tf.keras.regularizers.l2(0.001)))\n",
"model.add(tf.keras.layers.BatchNormalization())\n",
"model.add(tf.keras.layers.Dropout(0.5))\n",
"\n",
"\n",
"model.add(tf.keras.layers.Dense(num_classes,activation='softmax'))\n",
"\n",
"\n",
"model.compile(\n",
" optimizer=tf.keras.optimizers.Adam(learning_rate=3e-4),\n",
" loss='sparse_categorical_crossentropy',\n",
" metrics=['accuracy']\n",
")\n",
"\n",
"early_stopping=tf.keras.callbacks.EarlyStopping(\n",
" monitor='val_loss',\n",
" patience=7,\n",
" restore_best_weights=True,\n",
" verbose=1\n",
")\n",
"\n",
"\n",
"reduce_lr=tf.keras.callbacks.ReduceLROnPlateau(\n",
" monitor='val_loss',\n",
" factor=0.5,\n",
" patience=3,\n",
" min_lr=1e-7,\n",
" verbose=1\n",
")\n",
"\n",
"call_backs=[early_stopping,reduce_lr]\n",
"\n",
"history=model.fit(\n",
" train_dataset,\n",
" validation_data=validation_dataset,\n",
" epochs=50,\n",
" callbacks=call_backs,\n",
" verbose=1\n",
")\n",
"\n",
"print(\"\\n Training Complete\")\n",
"print(f\"Best Validation accuracy :{max(history.history['val_accuracy']):.4f}\")"
]
},
{
"cell_type": "markdown",
"id": "423633e8",
"metadata": {},
"source": [
"## Insights so far"
]
},
{
"cell_type": "markdown",
"id": "cb77a38d",
"metadata": {},
"source": [
"## Transfer Learning - MobileNetV2"
]
},
{
"cell_type": "code",
"execution_count": 30,
"id": "e738a856",
"metadata": {},
"outputs": [],
"source": [
"base_model=tf.keras.applications.MobileNetV2(\n",
" input_shape=(224,224,3),\n",
" include_top=False,\n",
" weights='imagenet'\n",
")"
]
},
{
"cell_type": "code",
"execution_count": 31,
"id": "a7c452b6",
"metadata": {},
"outputs": [],
"source": [
"base_model.trainable=False\n",
"\n",
"model=tf.keras.models.Sequential()\n",
"\n",
"model.add(tf.keras.layers.Input(shape=(224,224,3)))\n",
"model.add(data_argumentation)\n",
"model.add(tf.keras.layers.Rescaling(1.0/255))\n",
"\n",
"model.add(base_model)\n",
"model.add(tf.keras.layers.GlobalAveragePooling2D())\n",
"\n",
"model.add(tf.keras.layers.Dense(128,activation='relu'))\n",
"model.add(tf.keras.layers.Dense(num_classes,activation='softmax'))"
]
},
{
"cell_type": "code",
"execution_count": 32,
"id": "4f0c00d4",
"metadata": {},
"outputs": [],
"source": [
"model.compile(\n",
" optimizer='adam',\n",
" loss='sparse_categorical_crossentropy',\n",
" metrics=['accuracy']\n",
")"
]
},
{
"cell_type": "code",
"execution_count": 33,
"id": "e5b82a63",
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"Model: \"sequential_9\"\n",
"\n"
],
"text/plain": [
"\u001b[1mModel: \"sequential_9\"\u001b[0m\n"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
"┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n",
"┃ Layer (type) ┃ Output Shape ┃ Param # ┃\n",
"┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n",
"│ sequential_6 (Sequential) │ (None, 224, 224, 3) │ 0 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ rescaling_6 (Rescaling) │ (None, 224, 224, 3) │ 0 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ mobilenetv2_1.00_224 │ (None, 7, 7, 1280) │ 2,257,984 │\n",
"│ (Functional) │ │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ global_average_pooling2d │ (None, 1280) │ 0 │\n",
"│ (GlobalAveragePooling2D) │ │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ dense_14 (Dense) │ (None, 128) │ 163,968 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ dense_15 (Dense) │ (None, 6) │ 774 │\n",
"└─────────────────────────────────┴────────────────────────┴───────────────┘\n",
"\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",
"│ sequential_6 (\u001b[38;5;33mSequential\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m224\u001b[0m, \u001b[38;5;34m224\u001b[0m, \u001b[38;5;34m3\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ rescaling_6 (\u001b[38;5;33mRescaling\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m224\u001b[0m, \u001b[38;5;34m224\u001b[0m, \u001b[38;5;34m3\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ mobilenetv2_1.00_224 │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m7\u001b[0m, \u001b[38;5;34m7\u001b[0m, \u001b[38;5;34m1280\u001b[0m) │ \u001b[38;5;34m2,257,984\u001b[0m │\n",
"│ (\u001b[38;5;33mFunctional\u001b[0m) │ │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ global_average_pooling2d │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m1280\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │\n",
"│ (\u001b[38;5;33mGlobalAveragePooling2D\u001b[0m) │ │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ dense_14 (\u001b[38;5;33mDense\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m128\u001b[0m) │ \u001b[38;5;34m163,968\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ dense_15 (\u001b[38;5;33mDense\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m6\u001b[0m) │ \u001b[38;5;34m774\u001b[0m │\n",
"└─────────────────────────────────┴────────────────────────┴───────────────┘\n"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
" Total params: 2,422,726 (9.24 MB)\n",
"\n"
],
"text/plain": [
"\u001b[1m Total params: \u001b[0m\u001b[38;5;34m2,422,726\u001b[0m (9.24 MB)\n"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
" Trainable params: 164,742 (643.52 KB)\n",
"\n"
],
"text/plain": [
"\u001b[1m Trainable params: \u001b[0m\u001b[38;5;34m164,742\u001b[0m (643.52 KB)\n"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
" Non-trainable params: 2,257,984 (8.61 MB)\n",
"\n"
],
"text/plain": [
"\u001b[1m Non-trainable params: \u001b[0m\u001b[38;5;34m2,257,984\u001b[0m (8.61 MB)\n"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"model.summary()"
]
},
{
"cell_type": "code",
"execution_count": 34,
"id": "30628ca8",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Epoch 1/20\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m19s\u001b[0m 506ms/step - accuracy: 0.4802 - loss: 1.5135 - val_accuracy: 0.5876 - val_loss: 1.1383\n",
"Epoch 2/20\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m10s\u001b[0m 404ms/step - accuracy: 0.6709 - loss: 0.9045 - val_accuracy: 0.6497 - val_loss: 0.9297\n",
"Epoch 3/20\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m10s\u001b[0m 439ms/step - accuracy: 0.7472 - loss: 0.7275 - val_accuracy: 0.6554 - val_loss: 0.9454\n",
"Epoch 4/20\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m10s\u001b[0m 443ms/step - accuracy: 0.7514 - loss: 0.6609 - val_accuracy: 0.6497 - val_loss: 0.9596\n",
"Epoch 5/20\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m10s\u001b[0m 408ms/step - accuracy: 0.7895 - loss: 0.5747 - val_accuracy: 0.7119 - val_loss: 0.7839\n",
"Epoch 6/20\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m10s\u001b[0m 425ms/step - accuracy: 0.8023 - loss: 0.5816 - val_accuracy: 0.6497 - val_loss: 0.9661\n",
"Epoch 7/20\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m10s\u001b[0m 391ms/step - accuracy: 0.8404 - loss: 0.4513 - val_accuracy: 0.6723 - val_loss: 0.8813\n",
"Epoch 8/20\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m10s\u001b[0m 416ms/step - accuracy: 0.7867 - loss: 0.5593 - val_accuracy: 0.6893 - val_loss: 0.8278\n",
"Epoch 9/20\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m10s\u001b[0m 401ms/step - accuracy: 0.8291 - loss: 0.4717 - val_accuracy: 0.7345 - val_loss: 0.7541\n",
"Epoch 10/20\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 395ms/step - accuracy: 0.8517 - loss: 0.4349 - val_accuracy: 0.6723 - val_loss: 0.9188\n",
"Epoch 11/20\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m10s\u001b[0m 403ms/step - accuracy: 0.8545 - loss: 0.3900 - val_accuracy: 0.6949 - val_loss: 0.8123\n",
"Epoch 12/20\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 388ms/step - accuracy: 0.8686 - loss: 0.4226 - val_accuracy: 0.7175 - val_loss: 0.8446\n",
"Epoch 13/20\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m10s\u001b[0m 418ms/step - accuracy: 0.8573 - loss: 0.3874 - val_accuracy: 0.6893 - val_loss: 0.9091\n",
"Epoch 14/20\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m10s\u001b[0m 412ms/step - accuracy: 0.8927 - loss: 0.3323 - val_accuracy: 0.6271 - val_loss: 1.0940\n",
"Epoch 15/20\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 400ms/step - accuracy: 0.8743 - loss: 0.3465 - val_accuracy: 0.7345 - val_loss: 0.7991\n",
"Epoch 16/20\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m10s\u001b[0m 398ms/step - accuracy: 0.9096 - loss: 0.3085 - val_accuracy: 0.6667 - val_loss: 0.8792\n",
"Epoch 17/20\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 403ms/step - accuracy: 0.8630 - loss: 0.3604 - val_accuracy: 0.6780 - val_loss: 1.0453\n",
"Epoch 18/20\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m10s\u001b[0m 419ms/step - accuracy: 0.8814 - loss: 0.3261 - val_accuracy: 0.6780 - val_loss: 0.9043\n",
"Epoch 19/20\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m10s\u001b[0m 400ms/step - accuracy: 0.8941 - loss: 0.3041 - val_accuracy: 0.6949 - val_loss: 0.8192\n",
"Epoch 20/20\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m10s\u001b[0m 428ms/step - accuracy: 0.9195 - loss: 0.2597 - val_accuracy: 0.7006 - val_loss: 0.8011\n"
]
}
],
"source": [
"history=model.fit(\n",
" train_dataset,\n",
" validation_data=validation_dataset,\n",
" epochs=20\n",
")"
]
},
{
"cell_type": "code",
"execution_count": 35,
"id": "306964eb",
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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",
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"train_accuracy=history.history['accuracy']\n",
"val_accuracy=history.history['val_accuracy']\n",
"epochs=range(1,21)\n",
"\n",
"plt.figure(figsize=(10,5))\n",
"plt.plot(epochs,train_accuracy,label='Training Accuracy ',marker='o')\n",
"plt.plot(epochs,val_accuracy,label='Validation Accuracy',marker='o')\n",
"plt.title(\"Training vs Validation Accuracy\")\n",
"plt.xlabel(\"Epochs\")\n",
"plt.ylabel(\"Accuracy\")\n",
"plt.legend()\n",
"plt.show()"
]
},
{
"cell_type": "code",
"execution_count": 36,
"id": "870b6dce",
"metadata": {},
"outputs": [
{
"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": [
"model.save(\"../Models/MobileNetV2.h5\")"
]
},
{
"cell_type": "markdown",
"id": "b48aa932",
"metadata": {},
"source": [
"## EfficientNet"
]
},
{
"cell_type": "code",
"execution_count": 38,
"id": "6ea0e347",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Found 885 files belonging to 6 classes.\n",
"Using 708 files for training.\n"
]
}
],
"source": [
"train_dataset=tf.keras.utils.image_dataset_from_directory(\n",
" DATASET_DIR,\n",
" validation_split=0.2,\n",
" subset='training',\n",
" image_size=(IMG_WIDTH,IMG_HEIGHT),\n",
" batch_size=BATCH_SIZE,\n",
" seed=SEED\n",
")"
]
},
{
"cell_type": "code",
"execution_count": 40,
"id": "f5a15a59",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Found 885 files belonging to 6 classes.\n",
"Using 177 files for validation.\n"
]
}
],
"source": [
"validation_dataset=tf.keras.utils.image_dataset_from_directory(\n",
" DATASET_DIR,\n",
" validation_split=0.2,\n",
" subset=\"validation\",\n",
" image_size=(IMG_HEIGHT,IMG_WIDTH),\n",
" batch_size=BATCH_SIZE,\n",
" seed = SEED\n",
")"
]
},
{
"cell_type": "code",
"execution_count": 41,
"id": "5608f23d",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"['Bird-drop',\n",
" 'Clean',\n",
" 'Dusty',\n",
" 'Electrical-damage',\n",
" 'Physical-Damage',\n",
" 'Snow-Covered']"
]
},
"execution_count": 41,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"class_names"
]
},
{
"cell_type": "code",
"execution_count": 42,
"id": "a3a0fc62",
"metadata": {},
"outputs": [],
"source": [
"num_classes=len(class_names)"
]
},
{
"cell_type": "code",
"execution_count": 43,
"id": "72fc4d7e",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"6"
]
},
"execution_count": 43,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"num_classes"
]
},
{
"cell_type": "code",
"execution_count": 44,
"id": "04e3193a",
"metadata": {},
"outputs": [],
"source": [
"class_counts={}\n",
"total_images=0\n",
"\n",
"for class_name in class_names:\n",
" class_path=os.path.join(DATASET_DIR,class_name)\n",
" count=len(os.listdir(class_path))\n",
" class_counts[class_name]=count\n",
" total_images+=count"
]
},
{
"cell_type": "code",
"execution_count": 49,
"id": "971f6f03",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"class Weights {0: 0.7530224525043178, 1: 0.7491408934707904, 2: 0.7649122807017544, 3: 1.4110032362459546, 4: 2.106280193236715, 5: 1.1815718157181572}\n"
]
}
],
"source": [
"class_weights={}\n",
"\n",
"for index,class_name in enumerate(class_names):\n",
" class_weights[index]=total_images/(num_classes*class_counts[class_name])\n",
"\n",
"print(\"class Weights\",class_weights)"
]
},
{
"cell_type": "code",
"execution_count": 50,
"id": "57ebda02",
"metadata": {},
"outputs": [],
"source": [
"data_argumentation=tf.keras.Sequential()\n",
"data_argumentation.add(tf.keras.layers.RandomFlip('horizontal'))\n",
"data_argumentation.add(tf.keras.layers.RandomRotation(0.1))\n",
"data_argumentation.add(tf.keras.layers.RandomZoom(0.1))"
]
},
{
"cell_type": "code",
"execution_count": 51,
"id": "5db39d0e",
"metadata": {},
"outputs": [],
"source": [
"train_images=[]\n",
"train_labels=[]"
]
},
{
"cell_type": "code",
"execution_count": 52,
"id": "fd9fd9e0",
"metadata": {},
"outputs": [],
"source": [
"for images ,labels in train_dataset:\n",
" images=tf.keras.applications.efficientnet.preprocess_input(images)\n",
" train_images.append(images)\n",
" train_labels.append(labels)"
]
},
{
"cell_type": "code",
"execution_count": 53,
"id": "34a98406",
"metadata": {},
"outputs": [],
"source": [
"train_images=tf.concat(train_images,axis=0)\n",
"train_labels=tf.concat(train_labels,axis=0)"
]
},
{
"cell_type": "code",
"execution_count": 56,
"id": "a7b03e31",
"metadata": {},
"outputs": [],
"source": [
"validation_images=[]\n",
"validation_labels=[]"
]
},
{
"cell_type": "code",
"execution_count": 57,
"id": "e0fd0944",
"metadata": {},
"outputs": [],
"source": [
"for images ,labels in validation_dataset:\n",
" images=tf.keras.applications.efficientnet.preprocess_input(images)\n",
" validation_images.append(images)\n",
" validation_labels.append(labels)"
]
},
{
"cell_type": "code",
"execution_count": 59,
"id": "5a20c06c",
"metadata": {},
"outputs": [],
"source": [
"validation_images=tf.concat(validation_images,axis=0)\n",
"validation_labels=tf.concat(validation_labels,axis=0)"
]
},
{
"cell_type": "code",
"execution_count": 66,
"id": "a37ded76",
"metadata": {},
"outputs": [],
"source": [
"base_model=tf.keras.applications.EfficientNetB0(\n",
" input_shape=(224,224,3),\n",
" include_top=False,\n",
" weights='imagenet'\n",
")"
]
},
{
"cell_type": "code",
"execution_count": 67,
"id": "74d1d2e7",
"metadata": {},
"outputs": [],
"source": [
"base_model.trainable=False"
]
},
{
"cell_type": "code",
"execution_count": 68,
"id": "c641d7d0",
"metadata": {},
"outputs": [],
"source": [
"model=tf.keras.models.Sequential()\n",
"\n",
"model.add(tf.keras.layers.Input(shape=(224,224,3)))\n",
"model.add(data_argumentation)\n",
"\n",
"model.add(base_model)\n",
"model.add(tf.keras.layers.GlobalAveragePooling2D())\n",
"\n",
"model.add(tf.keras.layers.Dense(128,activation='relu'))\n",
"model.add(tf.keras.layers.Dense(num_classes,activation='softmax'))"
]
},
{
"cell_type": "code",
"execution_count": 69,
"id": "b5883dd0",
"metadata": {},
"outputs": [],
"source": [
"model.compile(\n",
" optimizer='adam',\n",
" loss='sparse_categorical_crossentropy',\n",
" metrics=['accuracy']\n",
")"
]
},
{
"cell_type": "code",
"execution_count": 70,
"id": "c56f65c3",
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"Model: \"sequential_13\"\n",
"\n"
],
"text/plain": [
"\u001b[1mModel: \"sequential_13\"\u001b[0m\n"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
"┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n",
"┃ Layer (type) ┃ Output Shape ┃ Param # ┃\n",
"┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n",
"│ sequential_10 (Sequential) │ (None, 224, 224, 3) │ 0 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ efficientnetb0 (Functional) │ (None, 7, 7, 1280) │ 4,049,571 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ global_average_pooling2d_1 │ (None, 1280) │ 0 │\n",
"│ (GlobalAveragePooling2D) │ │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ dense_16 (Dense) │ (None, 128) │ 163,968 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ dense_17 (Dense) │ (None, 6) │ 774 │\n",
"└─────────────────────────────────┴────────────────────────┴───────────────┘\n",
"\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",
"│ sequential_10 (\u001b[38;5;33mSequential\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m224\u001b[0m, \u001b[38;5;34m224\u001b[0m, \u001b[38;5;34m3\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ efficientnetb0 (\u001b[38;5;33mFunctional\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m7\u001b[0m, \u001b[38;5;34m7\u001b[0m, \u001b[38;5;34m1280\u001b[0m) │ \u001b[38;5;34m4,049,571\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ global_average_pooling2d_1 │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m1280\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │\n",
"│ (\u001b[38;5;33mGlobalAveragePooling2D\u001b[0m) │ │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ dense_16 (\u001b[38;5;33mDense\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m128\u001b[0m) │ \u001b[38;5;34m163,968\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ dense_17 (\u001b[38;5;33mDense\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m6\u001b[0m) │ \u001b[38;5;34m774\u001b[0m │\n",
"└─────────────────────────────────┴────────────────────────┴───────────────┘\n"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
" Total params: 4,214,313 (16.08 MB)\n",
"\n"
],
"text/plain": [
"\u001b[1m Total params: \u001b[0m\u001b[38;5;34m4,214,313\u001b[0m (16.08 MB)\n"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
" Trainable params: 164,742 (643.52 KB)\n",
"\n"
],
"text/plain": [
"\u001b[1m Trainable params: \u001b[0m\u001b[38;5;34m164,742\u001b[0m (643.52 KB)\n"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
" Non-trainable params: 4,049,571 (15.45 MB)\n",
"\n"
],
"text/plain": [
"\u001b[1m Non-trainable params: \u001b[0m\u001b[38;5;34m4,049,571\u001b[0m (15.45 MB)\n"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"model.summary()"
]
},
{
"cell_type": "code",
"execution_count": 71,
"id": "dbc0e9e8",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Epoch 1/15\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m32s\u001b[0m 760ms/step - accuracy: 0.5000 - loss: 1.3037 - val_accuracy: 0.6045 - val_loss: 0.9725\n",
"Epoch 2/15\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 526ms/step - accuracy: 0.7444 - loss: 0.7406 - val_accuracy: 0.7175 - val_loss: 0.7789\n",
"Epoch 3/15\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 499ms/step - accuracy: 0.8037 - loss: 0.5808 - val_accuracy: 0.7458 - val_loss: 0.7599\n",
"Epoch 4/15\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 504ms/step - accuracy: 0.8291 - loss: 0.4990 - val_accuracy: 0.7571 - val_loss: 0.7019\n",
"Epoch 5/15\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 477ms/step - accuracy: 0.8588 - loss: 0.4562 - val_accuracy: 0.7401 - val_loss: 0.7037\n",
"Epoch 6/15\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 497ms/step - accuracy: 0.8715 - loss: 0.4018 - val_accuracy: 0.7627 - val_loss: 0.6420\n",
"Epoch 7/15\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m13s\u001b[0m 548ms/step - accuracy: 0.8912 - loss: 0.3217 - val_accuracy: 0.7740 - val_loss: 0.6264\n",
"Epoch 8/15\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m13s\u001b[0m 545ms/step - accuracy: 0.9124 - loss: 0.2979 - val_accuracy: 0.7853 - val_loss: 0.6294\n",
"Epoch 9/15\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 516ms/step - accuracy: 0.8884 - loss: 0.3280 - val_accuracy: 0.7910 - val_loss: 0.6832\n",
"Epoch 10/15\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 488ms/step - accuracy: 0.9251 - loss: 0.2557 - val_accuracy: 0.7853 - val_loss: 0.6152\n",
"Epoch 11/15\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 496ms/step - accuracy: 0.8983 - loss: 0.2809 - val_accuracy: 0.8136 - val_loss: 0.5492\n",
"Epoch 12/15\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 491ms/step - accuracy: 0.9534 - loss: 0.2055 - val_accuracy: 0.8192 - val_loss: 0.5921\n",
"Epoch 13/15\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 489ms/step - accuracy: 0.9322 - loss: 0.2121 - val_accuracy: 0.8079 - val_loss: 0.6947\n",
"Epoch 14/15\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 498ms/step - accuracy: 0.9435 - loss: 0.1934 - val_accuracy: 0.8023 - val_loss: 0.6609\n",
"Epoch 15/15\n",
"\u001b[1m23/23\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 506ms/step - accuracy: 0.9477 - loss: 0.1881 - val_accuracy: 0.8079 - val_loss: 0.6548\n"
]
},
{
"data": {
"text/plain": [
""
]
},
"execution_count": 71,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"model.fit(\n",
" train_dataset,\n",
" validation_data=validation_dataset,\n",
" epochs=15\n",
")"
]
},
{
"cell_type": "markdown",
"id": "4918c761",
"metadata": {},
"source": [
"## Hyper Parameter Optimization"
]
},
{
"cell_type": "code",
"execution_count": 74,
"id": "a070c4f4",
"metadata": {},
"outputs": [],
"source": [
"import keras_tuner as kt"
]
},
{
"cell_type": "code",
"execution_count": 79,
"id": "a4548eeb",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Trial 20 Complete [00h 02m 21s]\n",
"val_accuracy: 0.7570621371269226\n",
"\n",
"Best val_accuracy So Far: 0.8305084705352783\n",
"Total elapsed time: 01h 14m 34s\n",
"Best Hyperparameters:\n",
"{'rotation_factor': 0.1, 'zoom_factor': 0.1, 'dropout_rate': 0.1, 'dense_units': 128, 'learning_rate': 0.0018290793529110935}\n",
"\u001b[1m6/6\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 340ms/step - accuracy: 0.8305 - loss: 0.5819\n",
"Best Validation Accuracy :0.8305\n"
]
}
],
"source": [
"def build_model(hp):\n",
" base_model=tf.keras.applications.EfficientNetB0(\n",
" input_shape=(224,224,3),\n",
" include_top=False,\n",
" weights='imagenet'\n",
" )\n",
" base_model.trainable=False\n",
"\n",
" model=tf.keras.models.Sequential(\n",
" [\n",
" tf.keras.layers.RandomFlip('horizontal'),\n",
" tf.keras.layers.RandomRotation(hp.Float(\"rotation_factor\",min_value=0.05,max_value=0.3,step=0.05)),\n",
" tf.keras.layers.RandomZoom(hp.Float(\"zoom_factor\",min_value=0.05,max_value=0.3,step=0.05)),\n",
" base_model,\n",
" tf.keras.layers.GlobalAveragePooling2D(),\n",
" tf.keras.layers.Dropout(hp.Float('dropout_rate',min_value=0.0,max_value=0.5,step=0.1)),\n",
" tf.keras.layers.Dense(hp.Int('dense_units',min_value=64,max_value=512,step=64),activation='relu'),\n",
" tf.keras.layers.Dense(num_classes,activation='softmax')\n",
" ]\n",
" )\n",
" model.compile(\n",
" optimizer=tf.keras.optimizers.Adam(hp.Float(\"learning_rate\",min_value=1e-4,max_value=1e-2,sampling='log')),\n",
" loss='sparse_categorical_crossentropy',\n",
" metrics=['accuracy']\n",
"\n",
" )\n",
" return model\n",
"\n",
"tuner=kt.RandomSearch(\n",
" build_model,\n",
" objective='val_accuracy',\n",
" max_trials=20,\n",
" executions_per_trial=1,\n",
" directory='kt-dir',\n",
" project_name='efficientnet_tune',\n",
" overwrite=True\n",
")\n",
"\n",
"\n",
"early_stopping=tf.keras.callbacks.EarlyStopping(monitor='val_accuracy',patience=5,restore_best_weights=True)\n",
"\n",
"\n",
"tuner.search(\n",
" train_dataset,\n",
" validation_data=validation_dataset,\n",
" epochs=20,\n",
" callbacks=[early_stopping],\n",
" class_weight=class_weights\n",
")\n",
"\n",
"best_hps=tuner.get_best_hyperparameters(num_trials=1)[0]\n",
"print(\"Best Hyperparameters:\")\n",
"print(best_hps.values)\n",
"\n",
"best_model=tuner.get_best_models(num_models=1)[0]\n",
"\n",
"val_loss,val_acc=best_model.evaluate(validation_dataset)\n",
"print(f\"Best Validation Accuracy :{val_acc:.4f}\")"
]
},
{
"cell_type": "code",
"execution_count": 80,
"id": "9512ce43",
"metadata": {},
"outputs": [
{
"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": [
"best_model.save(\"../Models/effnet_finetune.h5\")"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "9c2bda40",
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "ai-ml",
"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": "3.10.20"
}
},
"nbformat": 4,
"nbformat_minor": 5
}