Deep Learning (DL) · Intermediate level

Intermediate level · Module 3 of 6 · 16 guided hours

Building Image and Text Data Sets

Collect and label data well, augment it sensibly and prepare text for deep learning models.

Learning objectives

  • Plan data collection and write labelling guidelines
  • Measure and improve label quality and avoid split leakage
  • Apply image augmentation and text preprocessing appropriately

Topics

  • data collection plans
  • labelling guidelines and tools
  • label quality checks
  • near-duplicates and split leakage
  • image augmentation strategies
  • text tokenisation and subwords

Lessons

  1. Lesson 1 Collecting and Labelling Data

    Plan what data to collect and write labelling guidelines that produce consistent labels.

    coverage of conditionsclass balancelabelling guidelineslabelling tools

  2. Lesson 2 Label Quality and Leakage-Free Splits

    Measure label agreement, find label errors and split data so evaluation is honest.

    inter-annotator agreementlabel noisenear-duplicate imagesgroup-based splitting

  3. Lesson 3 Augmentation and Text Preparation

    Choose augmentations that mimic real variation, and prepare text with modern tokenisers.

    realistic augmentationaugmentation that breaks labelssubword tokenisationsequence length

Practical activity

Collect and label medicine package photos as a class, following a labelling guide, then check label quality and build leakage-free splits.

Expected outcomes

  • High-quality, well-documented data sets
  • Leakage-free evaluation data

Assessment

Data set review: guidelines, quality metrics, split method and documentation.

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.