Deep Learning (DL)

Level 1 of 3

Beginner Level

Prerequisites

Python Beginner and Machine Learning Beginner levels, or equivalent: Python functions and lists, training and evaluating a model, and the idea of overfitting.

Duration

4 weeks · 52 guided hours · 5 modules. Taught live by an instructor — online or in person.

By the end of this level you can

  • Explain how neural networks compute predictions and learn from errors
  • Build and train a small image classifier with a convolutional network in PyTorch
  • Describe how sequence models process text and time series
  • Adapt a pre-trained model to a new task with transfer learning
  • Train networks reliably by choosing losses, optimisers, learning rates and regularisation

Modules

  1. 1Module 1 · 10 h · 3 lessons · free taster lessonNeural Network Foundations
  2. 2Module 2 · 11 h · 3 lessons · free taster lessonConvolutional Networks for Images
  3. 3Module 3 · 10 h · 3 lessons · free taster lessonSequence Models for Text and Time
  4. 4Module 4 · 11 h · 3 lessons · free taster lessonTransfer Learning with Pre-trained Models
  5. 5Module 5 · 10 h · 3 lessons · free taster lessonTraining Neural Networks Well
  6. Free course quizQuick Quiz — Deep Learning Basics

Beginner level project

Leaf Disease Classifier with Transfer Learning

Fine-tune a pre-trained image model on a provided open data set of crop leaf photos to recognise healthy and diseased leaves, and report its accuracy and limits for use by agricultural extension workers.

Deliverables

  • A notebook that loads the data, fine-tunes a pre-trained CNN and evaluates it on a held-out test set
  • Training and validation curves with a short diagnosis
  • A confusion matrix and examples of correct and incorrect predictions
  • A one-page note on how and how not to use the model in the field

Assessed on

  • The model clearly beats a trivial baseline on the untouched test set
  • Training choices (learning rate, epochs, augmentation) are justified with evidence
  • Errors are analysed and limits — such as photos unlike the training data — are stated
  • The notebook runs top to bottom on a free cloud GPU or a CPU

All levels are taught live by an instructor — online or in person — see registration options.