7. Write a program to enable pre-train models to classify a given image dataset.
📁 Folder Structure for Pre-Trained Image Classification
project_folder/ ← Your main project folder
├── main.py ← Python script
└── data/ ← Dataset folder
├── train/ ← Training images
│ ├── cat/ ← Class 1 training images
│ │ ├── cat1.jpg
│ │ ├── cat2.jpg
│ │ └── ...
│ └── dog/ ← Class 2 training images
│ ├── dog1.jpg
│ ├── dog2.jpg
│ └── ...
└── val/ ← Validation images
├── cat/ ← Class 1 validation images
│ ├── cat3.jpg
│ └── ...
└── dog/ ← Class 2 validation images
├── dog3.jpg
└── ...✅ Install Required Packages
pip install tensorflowPROGRAM:
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from tensorflow.keras.models import Model
from tensorflow.keras.layers import Dense, GlobalAveragePooling2D
from tensorflow.keras.applications import ResNet50
from tensorflow.keras.applications.resnet50 import preprocess_input
# 1) Dataset paths adjust as per requirement
train_dir = "data/train"
val_dir = "data/val"
img_size = (224, 224)
batch_size = 32
# 2) Data generators
train_gen = ImageDataGenerator(preprocessing_function=preprocess_input).flow_from_directory(
train_dir, target_size=img_size, batch_size=batch_size, class_mode='categorical'
)
val_gen = ImageDataGenerator(preprocessing_function=preprocess_input).flow_from_directory(
val_dir, target_size=img_size, batch_size=batch_size, class_mode='categorical'
)
num_classes = train_gen.num_classes
# 3) Pre-trained model (ResNet50) without top layer
base_model = ResNet50(weights='imagenet', include_top=False, input_shape=(224,224,3))
base_model.trainable = False # Freeze base model
# 4) Add classification head
x = base_model.output
x = GlobalAveragePooling2D()(x)
x = Dense(128, activation='relu')(x)
predictions = Dense(num_classes, activation='softmax')(x)
model = Model(inputs=base_model.input, outputs=predictions)
# 5) Compile model
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])
# 6) Train model
model.fit(train_gen, validation_data=val_gen, epochs=5)
# 7) Evaluate model
loss, acc = model.evaluate(val_gen)
print(f"Validation Loss: {loss:.4f}, Validation Accuracy: {acc:.4f}")OUTPUT:
Note: The results you see (accuracy and loss) will depend on the images you use for training and validation. Using more or higher-quality images may change the output.
Found 18 images belonging to 2 classes.
Found 2 images belonging to 2 classes.
Epoch 1/5
1/1 [==============================] - 5s 5s/step - loss: 0.5678 - accuracy: 0.7222 - val_loss: 0.0982 - val_accuracy: 1.0000
Epoch 2/5
1/1 [==============================] - 1s 935ms/step - loss: 0.0476 - accuracy: 1.0000 - val_loss: 0.0172 - val_accuracy: 1.0000
Epoch 3/5
1/1 [==============================] - 1s 918ms/step - loss: 0.0094 - accuracy: 1.0000 - val_loss: 0.0040 - val_accuracy: 1.0000
Epoch 4/5
1/1 [==============================] - 1s 927ms/step - loss: 0.0031 - accuracy: 1.0000 - val_loss: 0.0011 - val_accuracy: 1.0000
Epoch 5/5
1/1 [==============================] - 1s 1s/step - loss: 0.0012 - accuracy: 1.0000 - val_loss: 3.5405e-04 - val_accuracy: 1.0000
1/1 [==============================] - 0s 126ms/step - loss: 3.5405e-04 - accuracy: 1.0000
Validation Loss: 0.0004, Validation Accuracy: 1.0000