GPMS TRANSPORTAL APPLIED AI COURSES

• This course covers the best practices for quality assurance (QA) in AI projects, specifically for computer vision applications. Students will learn how to test AI models, ensure the quality of input data, handle edge cases, and validate outcomes to guarantee that computer vision systems meet the desired business requirements and standards. The course also covers methods to deal with issues like overfitting, model bias, and performance issues in real world applications.
• Level: Intermediate
• Prerequisites: Basic understanding of AI and computer vision, no coding required.
• Assessments: Weekly Micro Assessments, Final Test and Evaluation.
• Job Roles Applicability: QA Engineer, AI/Computer Vision QA Specialist, Technical Tester, Data Validation Analyst.

---

Week 1 - Introduction to Quality Assurance in AI

Learning Outcome

Understand the core principles of quality assurance in AI, focusing on computer vision systems.

1.1 Importance of QA in AI projects, particularly in computer vision.
1.2 Key QA challenges in computer vision (data quality, model bias, edge cases).
1.3 Role of QA in the AI development lifecycle.

Practical Component
Introduction to QA tools for AI, reviewing sample computer vision systems for initial QA checks.

---

Week 2 - Understanding Data Quality and Preprocessing for Computer Vision

Learning Outcome

Learn the role of data in QA processes for computer vision and how to ensure the data is clean, labeled correctly, and representative.

2.1 Impact of data quality on model performance.
2.2 Techniques for labeling, annotating, and preprocessing data in computer vision tasks.
2.3 Handling imbalanced data and noisy inputs.

Practical Component
Review and clean a dataset, ensuring it meets the quality standards for a computer vision model.

---

Week 3 - Testing Computer Vision Models for Accuracy and Reliability

Learning Outcome

Explore strategies for testing computer vision models to ensure they meet business and technical objectives.

3.1 Understanding accuracy metrics for computer vision models (precision, recall, F1 score).
3.2 Testing model performance on validation and test datasets.
3.3 Evaluating reliability under various conditions (e.g., different lighting, angle, object variations).

Practical Componenet
Run tests on a computer vision model to assess its accuracy and robustness in a given environment.

---

Week 4 - Handling Model Bias and Ensuring Fairness in Computer Vision

Learning Outcome

Understand how to identify and mitigate biases in computer vision models and ensure fairness in AI applications.

4.1 What is model bias, and how does it affect computer vision?
4.2 Techniques to reduce bias and ensure model fairness.
4.3 Case studies of biased AI models and their implications.

Practical Componenet
Perform a fairness audit on a computer vision model to identify and address bias issues.

---

Week 5 - Edge Case Testing and Real-world Scenarios in Computer Vision

Learning Outcome

Learn how to design tests that account for edge cases and ensure that computer vision models work well in real-world scenarios

5.1 Identifying and testing for edge cases (e.g., unusual object shapes, occlusions, environmental factors).
5.2 Ensuring robustness against changes in input data.
5.3 Stress testing and scalability considerations in computer vision systems.

Practical Component
Create a set of edge cases and simulate real-world conditions to test a computer vision model's robustness.

---

Week 6 - Final Project and Assessment

Learning Outcome

Apply QA principles to a complete computer vision project, ensuring all aspects from data to model testing are thoroughly evaluated.

6.1 Conduct end-to-end QA on a computer vision project.
6.2 Final evaluations based on project robustness, performance, and fairness.
6.3 Presentation and peer review of final QA reports.

Practical Component
Complete a QA assessment for a computer vision model and present findings, highlighting key issues and proposed fixes.

---