Postgraduate Diploma in Unbalanced Datasets and VRU Metrics
Online 800-hour Postgraduate Diploma in Unbalanced Datasets and VRU Metrics. 8 modules with Regulation (EU) 2024/1689 (AI Act), ISO/PAS 8800:2024 and PyTorch. Assigned tutor and open enrolment.
Take the controls.
Run a real engineering trade-off from this field. Move the sliders, watch the model react in real time and beat the challenge.
What you will be able to do
Diagnose class and scenario imbalance in VRU datasets.
Plan data collection and curation strategies.
Manage annotation and its quality control.
Apply rebalancing techniques and synthetic data.
Select metrics suited to unbalanced data.
Evaluate performance across subgroups and its fairness.
Document datasets in line with AI and data regulations.
Who it's for
- Perception data engineer looking to specialise in Unbalanced Datasets and VRU Metrics.
- Data scientist looking to specialise in Unbalanced Datasets and VRU Metrics.
- Machine learning engineer looking to specialise in Unbalanced Datasets and VRU Metrics.
- Data quality manager looking to specialise in Unbalanced Datasets and VRU Metrics.
- ADAS validation engineer looking to specialise in Unbalanced Datasets and VRU Metrics.
- AI ethics specialist looking to specialise in Unbalanced Datasets and VRU Metrics.
- Graduates in engineering or sciences seeking to enter the automotive sector.
What you'll work with
- Regulation (EU) 2024/1689 (AI Act)
- ISO/PAS 8800:2024
- ISO 21448:2022 (SOTIF)
- Regulation (EU) 2016/679 (GDPR)
- ISO/IEC 5259-2:2024
- ISO/IEC 25012:2008
- ISO/IEC TR 24027:2021
8 modules · 800 h
1.1Class distribution in driving datasets1.2Long tail of rare categories1.3Imbalance of scenarios and conditions1.4Biases by age, height and clothing1.5Impact on system safety1.6Rare classes: wheelchairs and pushchairs1.7Exploratory analysis of labels1.8Documentation with datasheets1.9Geographical imbalance between cities1.10Practical case: audit of a public dataset
2.1Designing acquisition campaigns2.2Mining rare scenarios from the fleet2.3Active learning for data selection2.4Similarity search with embeddings2.5Deduplication and cleaning2.6Anonymisation of faces and number plates2.7Licensing and data governance2.8Dataset versioning2.9Capture triggers in the vehicle2.10Practice: active sample selection
3.1Annotation guidelines for VRUs3.2Occlusion, truncation and attributes3.33D annotation and tracking3.4Inter-annotator agreement3.5Detection of label errors3.6Semi-automatic labelling3.7Cost and productivity of annotation3.8Quality audit3.9Ambiguous labels and doubtful cases3.10Laboratory: measuring inter-annotator agreement
4.1Oversampling and undersampling4.2Sampling by class and by instance4.3Focal loss and weighted losses4.4Losses specific to long-tail distributions4.5Copy-paste and object-level augmentation4.6Synthetic VRU data4.7Avatars and pedestrian simulation4.8Risk of overfitting4.9Recalibrating the classifier after rebalancing4.10Laboratory: focal loss in a detector
5.1Precision, recall and the PR curve5.2mAP and its sensitivity to imbalance5.3Log-average miss rate5.4Metrics by size and distance5.5Risk-weighted metrics5.6Confidence calibration5.7Bootstrap confidence intervals5.8Safety-relevant errors5.9Selecting the operating threshold5.10Practice: stratified metrics in Python
6.1Defining evaluation subgroups6.2Performance disparity across groups6.3Applied fairness metrics6.4Sample size and significance6.5Analysis of critical scenarios6.6Coverage of the operational domain6.7Transparent evaluation reports6.8Mitigation of disparities6.9Hypothesis testing between subgroups6.10Practical case: fairness analysis of a detector
7.1EU Artificial Intelligence Act7.2Data quality according to ISO/PAS 88007.3Data requirements in SOTIF7.4GDPR and people on public roads7.5Data quality according to ISO/IEC 52597.6Technical documentation of the dataset7.7Traceability between data and requirements7.8Dataset lifecycle7.9Bias management according to ISO/IEC TR 240277.10Workshop: VRU data management plan
8.1Analysis of the starting dataset8.2Under-represented classes and scenarios8.3Acquisition and curation strategy8.4Annotation and quality plan8.5Rebalancing and synthetic data8.6Stratified metrics system8.7Fairness assessment8.8Dataset documentation8.9Dissertation and presentation8.10Final review with the tutor
How you'll learn
Fully 100% online training on the SEIUM campus, with downloadable material, recorded classes and an assigned technical tutor. No fixed timetables: students progress at their own pace within the intake, which makes it possible to combine the programme with work.
Each module combines theory, guided reading of the original standard and applied exercises. The working software is PyTorch, FiftyOne, CVAT, scikit-learn, DVC, in free, academic or evaluation versions. Practicals are based on real procedures: auditing the imbalance of a driving dataset, stratified evaluation of a pedestrian detector, and an active data selection campaign for VRUs. Assessment is continuous, module by module, plus the final integrative project.
Build it end to end
Your final project follows the real engineering workflow, step by step, reviewed by your tutor.
Diagnose class and scenario imbalance in VRU datasets.
Plan data collection and curation strategies.
Manage annotation and its quality control.
Apply rebalancing techniques and synthetic data.
Select metrics suited to unbalanced data.
Where this takes you
Requirements
- Prior technical training or professional experience in the sector. No university degree is required.
- Internet connection and a computer capable of running the programme's software.
- Sufficient level of Spanish to follow technical regulatory texts.
Ready to start?
Send your application today. Our admissions team replies within 72 hours with your study plan and funding options.
Frequently asked questions
The programme comprises 800 hours in total, spread over 6 to 8 months. That is an average of 10 to 15 hours per week, compatible with full-time work.
You will work with PyTorch, FiftyOne, CVAT, scikit-learn and DVC. Free, academic or evaluation versions are used, and the tutor will indicate the free alternative where one exists.
A SEIUM own diploma upon passing the continuous assessment and the integrative project. This is in-house training, not an official qualification from the Ministry.
Yes. Modules passed in a course or diploma in the same area can be credited towards the advanced programme. Please check the credit transfer policy when you enrol.
Related programmes
Postgraduate Diploma in Map-Based Prediction and Danger Hotspots
Online 800-hour Postgraduate Diploma in Map-Based Prediction and Danger Hotspots. 8 modules covering ISO 21448:2022 (SOTIF), ISO/PAS 8800:2024 and PyTorch. Assigned tutor and open enrolment.
Postgraduate Diploma in Robo-Taxi/Shuttle UX: Inclusion and Accessibility
Online 800 h Postgraduate Diploma in Robo-Taxi/Shuttle UX: Inclusion and Accessibility. 8 modules covering Directive (EU) 2019/882, EN 301 549 and Figma. Assigned tutor and open…
Postgraduate Diploma in SDV Platforms: Vehicle-Centric APIs, Sandbox and Permissions
Online 800-hour Postgraduate Diploma in SDV Platforms: Vehicle-Centric APIs, Sandbox and Permissions. 8 modules covering ISO/SAE 21434:2021, UN Regulation No. 155 and Eclipse Kuksa. Assigned tutor…
More in Autonomous, ADAS & Software-Defined Vehicle
All Autonomous, ADAS & Software-Defined Vehicle programmes → · How to apply · Fees & financing · Scholarships