Level 2 of 3
Intermediate Level
Prerequisites
Deep Learning Beginner level, or equivalent: training CNNs in PyTorch, transfer learning and basic regularisation.
Duration
8 weeks · 90 guided hours · 6 modules. Taught live by an instructor — online or in person.
By the end of this level you can
- Write clean, reusable PyTorch training code with datasets, loaders and checkpoints
- Use modern CNN architectures, including efficient models for mobile devices
- Prepare image and text data sets with quality labels and sound augmentation
- Train efficiently on GPUs and manage compute time and cost
- Evaluate deep learning models rigorously and diagnose their failures
Modules
- 1Module 1 · 15 h · 3 lessonsPyTorch in Depth
- 2Module 2 · 16 h · 3 lessonsModern CNN Architectures
- 3Module 3 · 16 h · 3 lessonsBuilding Image and Text Data Sets
- 4Module 4 · 13 h · 3 lessonsGPUs and Efficient Training
- 5Module 5 · 13 h · 3 lessonsEvaluating Deep Learning Models
- 6Module 6 · 17 h · 3 lessonsIntermediate Deep Learning Project
Intermediate level project
Medicine Package Recognition for Pharmacy Stock Checks
Build, train and evaluate an image classifier that recognises common medicine packages from phone photos (using a class-created or provided data set with no personal data), designed to run on a modest smartphone.
Deliverables
- A documented data set with a labelling guide and quality checks
- Reusable PyTorch training code with configuration, logging and checkpoints
- A comparison of at least two architectures, including one mobile-friendly model
- An evaluation report with error analysis and recommendations
Assessed on
- The data set is split without leakage (no near-duplicate photos across splits)
- Training code is reproducible and resumes from checkpoints
- The mobile-friendly model meets an agreed accuracy and size target
- Error analysis identifies the main failure types and proposes fixes
All levels are taught live by an instructor — online or in person — see registration options.