FIT3205 · Applied computer vision

Overview

In this unit you will develop applied competence in computer vision-related perception systems that perform reliably and ethically in real-world conditions. You will work across core vision tasks including classification, detection, segmentation, and representation learning, while addressing deployment realities such as dataset bias, spurious correlations, robustness to environmental change, safety and ethics in high‑impact contexts. Through hands-on labs you will build end-to-end vision pipelines, conduct systematic error analysis, and implement mitigations such as augmentation, calibration, and subgroup validation.

Offerings

The Handbook publishes no offerings for this unit.

Assessment

The Handbook publishes no assessment items for this unit yet.

Assessment details may change. Please refer to the assessment information in Moodle closer to the start of the teaching period.

Requisites

Learning outcomes

1. Design an end-to-end perception pipeline for a specified real-world application by defining task scope (e.g., classification, detection, segmentation), data requirements, evaluation strategy, and deployment constraints;
2. Select, apply and justify methods and tools for training, validating, optimising and deploying perception models, such as augmentation strategies, calibration techniques, subgroup validation and performance/latency tuning to meet real‑world constraints;
3. Implement and integrate vision models and preprocessing/postprocessing components into a reproducible pipeline, using appropriate software engineering practices to support experimentation and deployment;
4. Engineer and execute robust evaluation workflows, including systematic error analysis, calibration checks, and validation across relevant subgroups to characterise performance, failure modes, and uncertainty;
5. Diagnose and resolve complex performance and reliability issues by reasoning about data shift, spurious correlations, robustness to environmental variation, and trade-offs between accuracy, latency, and compute budgets, proposing justified mitigations such as augmentation or model optimisation/compression;
6. Assess and address ethical, safety, and societal risks when deploying perception systems, including bias and high-impact failure consequences, and articulate responsible limitations, monitoring needs, and governance controls appropriate to stakeholders and context.

Workload

No workload detail published.

Student experience, not official rules. Nothing here changes what the Handbook says.