Byte-Sized Deep Learning Series: Image Classification with CNN

Description:

How do computers "see" and recognize images?

This session will introduce you to Convolutional Neural Networks (CNNs): the deep learning architecture that powers technologies like facial recognition, autonomous vehicles, and medical image analysis.

You'll start by understanding why CNNs are essential for image tasks and how they cleverly use filters and feature maps to detect patterns like edges, shapes, and objects.

We’ll demystify the core concepts behind CNNs, explaining kernels, convolutions, pooling, and flattening in a simple and intuitive way.

You’ll use critical Keras layers such as Conv2D, MaxPooling2D, and Flatten, and learn how to design an efficient model architecture.

Duration: 90 mins

Course Code: BDT494

Learning Objectives:

After this course, you will be able to:

• Challenges with Artificial Neural Networks Dense layers with Images
• How do Convolutional Neural Networks (CNNs) work?
• Essential CNN building blocks in Keras
• Creating a CNN network for image classification

• Learners familiar with TensorFlow/Keras basics (Sequential models, training/fitting)

• Machine learning practitioners with basic experience building simple dense (fully connected) neural networks. Ideal for those ready to transition from tabular/text data to vision tasks.

Course Outline:

1. Why Convolutional Neural Networks (CNNs)
   • Challenges with Keras Dense layers for images
   • Local patterns and spatial dimensions
2. How CNNs work: Filters, Kernels, Feature Maps
   • Filters and Kernels: small windows sliding over input
   • Feature Maps: What the network “sees” at different stages
   • Convolution: Feature Extraction
   • Pooling: Down sampling feature maps
3. Essential CNN building blocks in Keras
   • Conv2D layer: setting filters, kernel sizes, strides
   • MaxPooling2D layer: down sampling feature maps
   • Flatten layer: Transitioning from 2D feature maps to dense output
4. CNN for image classification
   • Loading image dataset
   • Building a CNN model
   • Compiling the model
   • Training the model and evaluating test data
   • Hands-on: Performing image classification with CNN

Training material provided: Yes (Digital format)

Hands-on Lab: Instructions will be provided to install Jupyter notebook and other required python libraries. Students can opt to use ‘Google Colaboratory’ if they do not want to install these tools