Advanced level · Module 6 of 6 · 30 guided hours
Deep Learning Capstone
Design, train, compress, evaluate and document an offline field assistant model for a real-world task.
Learning objectives
- Plan a deep learning system around users and field constraints
- Train, compress and deploy it to meet targets
- Evaluate, document and present it responsibly
Topics
- user needs and targets
- data collection plan
- training and selection
- compression and on-device testing
- evaluation and model card
- presentation
Lessons
-
Lesson 1 Capstone Planning and Data
Define the task, users and targets, and prepare data responsibly.
user needsaccuracy, size and latency targetsdata plandata card
-
Lesson 2 Training and Model Selection
Train candidate models and select one that can meet all targets.
transfer learningexperiment trackingcandidate comparisonselection rationale
-
Lesson 3 Compression and On-Device Testing
Export, compress and run the model on the target device, measuring real performance.
exportquantisationon-device benchmarktarget check
-
Lesson 4 Evaluation and Documentation
Evaluate the deployed model on field-like data and document it fully.
field-like evaluationdisaggregated resultsmodel cardrisk assessment
-
Lesson 5 Capstone Presentation
Present the system with a live on-device demonstration and an honest account of its limits.
live demonstrationresults summarylimits and risksnext steps
Practical activity
Deliver the capstone through five reviewed milestones: plan and data, training, deployment, evaluation and documentation, presentation.
Expected outcomes
- A portfolio-quality deep learning system
- Experience of the full development cycle under real constraints
Assessment
The level project assessed against its criteria with an on-device demonstration.
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.