Using Pretrained Models

Pretrained models are deep learning models that have already been trained on large datasets. Instead of building a model from scratch, we can use these existing models and adapt them to our own tasks. This approach is widely used in transfer learning and is very effective for saving time and improving accuracy.

What are Pretrained Models?
Pretrained models are neural networks trained on large-scale datasets such as ImageNet. They have already learned useful features like edges, shapes, textures, and patterns. These learned features can be reused for new tasks.

Why Use Pretrained Models?

• Save training time and computing resources
• Improve accuracy with less data
• Reduce risk of overfitting
• Work well for complex deep learning tasks
• Provide strong baseline performance

How Pretrained Models Work

1. Feature Extraction

• Use the pretrained model as a fixed feature extractor
• Early layers are kept frozen
• Only new layers are trained

2. Fine-Tuning

• Unfreeze some layers of the pretrained model
• Train the model on new dataset
• Helps adapt to specific tasks

Popular Pretrained Models

• VGG16
• ResNet50
• InceptionV3
• MobileNet
• EfficientNet

Steps to Use Pretrained Models

Step 1: Load Pretrained Model

• Import model with pretrained weights

Step 2: Freeze Base Layers

• Prevent updating original learned features

Step 3: Add Custom Layers

• Add new layers for your specific problem

Step 4: Compile Model

• Choose optimizer, loss function, and metrics

Step 5: Train Model

• Train only new layers or fine-tune selected layers

Step 6: Evaluate Model

• Test performance on validation data

Example: Using Pretrained Model in Python (Keras)

from tensorflow.keras.applications import ResNet50
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Flatten# Load pretrained model
base_model = ResNet50(weights='imagenet', include_top=False, input_shape=(224, 224, 3))# Freeze base model layers
for layer in base_model.layers:
    layer.trainable = False# Build custom model
model = Sequential([
    base_model,
    Flatten(),
    Dense(128, activation='relu'),
    Dense(2, activation='softmax')
])model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])model.summary()

Advantages of Pretrained Models

• Faster development process
• High accuracy with small datasets
• Reduced computational cost
• Strong feature extraction capabilities

Limitations

• May not perfectly fit all custom tasks
• Requires careful fine-tuning
• Large model size may need high memory

Applications

• Image classification systems
• Object detection models
• Medical imaging analysis
• Face recognition systems
• Natural language processing tasks

Lesson Summary
Using pretrained models allows developers to leverage existing knowledge from large datasets. This makes deep learning faster, more efficient, and more accurate, especially when working with limited data or complex tasks.