Deep Learning (DL) · Intermediate level

Intermediate level · Module 2 of 6 · 16 guided hours

Modern CNN Architectures

Understand the architectures behind today's image models — deeper networks, batch normalisation, residual connections and efficient mobile models.

Learning objectives

  • Explain how deeper CNNs and batch normalisation improved accuracy
  • Explain residual connections and why they allow very deep networks
  • Choose efficient architectures for mobile and low-resource devices

Topics

  • deeper CNNs
  • batch normalisation
  • residual connections (ResNet)
  • depthwise separable convolutions
  • MobileNet and EfficientNet
  • accuracy vs size and speed

Lessons

  1. Lesson 1 Deeper Networks and Batch Normalisation

    See how stacking more layers improves features and how batch normalisation stabilises training.

    network depthreceptive fieldbatch normalisationtraining stability

  2. Lesson 2 Residual Connections

    Understand residual connections, the idea that made networks with dozens or hundreds of layers trainable.

    skip connectionresidual blockResNetgradient flow

  3. Lesson 3 Efficient Architectures for Mobile

    Use architectures designed for phones and low-power devices, and trade accuracy against size and speed.

    depthwise separable convolutionMobileNetparameters and FLOPslatency

Practical activity

Compare a ResNet and a MobileNet on the medicine package data for accuracy, model size and inference time on a laptop CPU.

Expected outcomes

  • Understanding of modern architectures
  • Architecture choices based on deployment constraints

Assessment

Comparison report marked on measurements and justified recommendations.

Taught live by an instructor — online or in person. Free course lessons are for self-study and do not lead to a certificate — see registration options.