Tutorial_05_CNN_in_Practice

Tutorial 5: CNN in Practice

This repository contains two tutorials showcasing practical applications of Convolutional Neural Networks (CNNs) using TensorFlow and PyTorch.

Tutorials Overview

1. CNN in Practice with TensorFlow

This tutorial demonstrates:

• Feature Extraction: Reusing a pre-trained CNN on a new dataset, a useful approach when working with small datasets.
• Visualization of Learned Concepts:
  • Intermediate ConvNet outputs.
  • Filters in ConvNets.
  • Heatmaps of class activation (GradCAM).

Note: The tutorial concludes with a feature extraction example, with fine-tuning covered in the next lecture.

2. CNN in Practice with PyTorch

This tutorial covers the same concepts as the TensorFlow tutorial, implemented using PyTorch:

• Feature Extraction: Leveraging pre-trained models for small datasets.
• Visualization of Learned Concepts:
  • Intermediate ConvNet outputs.
  • ConvNet filters.
  • GradCAM-based class activation heatmaps.

Acknowledgments

• François Chollet
• Tensorflow.org
• pytorch.org