Module 10: Applications and Ethics in Computer Vision

1. Learning Objectives

Explore impactful real-world applications of computer vision
Understand the societal implications and risks of vision-based systems
Evaluate ethical concerns in CV applications such as surveillance and facial recognition
Reflect on bias, fairness, and responsible AI development

2. Key Applications of Computer Vision

2.1 Optical Character Recognition (OCR)

Used for digitizing printed and handwritten documents
Applications: Scanned books, ID documents, bank cheques
Popular libraries: Tesseract, EasyOCR

2.2 Facial Recognition

Used in security, device unlocking, and user verification
Controversial use in law enforcement and mass surveillance
Bias against certain demographics is a major ethical concern

2.3 Medical Imaging

Assists in diagnosing diseases from X-rays, MRIs, CT scans
Early detection of cancer, tumors, and anomalies
Models must be explainable and carefully validated

2.4 Autonomous Vehicles

Uses cameras and LiDAR for lane detection, object tracking
Critical for decision-making in real-time environments
Failure can result in serious accidents or harm

3. Ethical Concerns in Computer Vision

3.1 Bias and Fairness

Training data often overrepresents certain demographics
Vision systems may fail on underrepresented groups
Example: Facial recognition having lower accuracy on darker-skinned faces

3.2 Privacy and Consent

Surveillance systems often operate without explicit user consent
Smart cameras track individuals in public/private spaces
Data protection and anonymity are essential considerations

3.3 Explainability

Deep learning systems are often black boxes
Important to provide interpretable results, especially in high-risk applications like healthcare

3.4 Deepfakes and Misinformation

GANs can create highly realistic fake videos and images
Misused to spread propaganda, fake news, or manipulate public opinion

4. Case Study: Amazon Rekognition

Amazon Rekognition is a facial analysis and recognition service.

Tested by law enforcement and governments
Criticized for racial bias and wrongful identifications
Led to bans and calls for regulation

Lesson: Commercial CV systems must be tested thoroughly and ethically before deployment.

5. Discussion: Responsible AI for Vision

Considerations for building ethical CV systems:

Diverse and representative training datasets
Regular auditing and fairness evaluations
User transparency and opt-in mechanisms
Regulations and policy frameworks for high-risk use cases

6. Assignment

Objective: Reflect on ethical challenges in computer vision applications.

Choose 1 application (e.g., facial recognition, OCR, medical AI)
Write a 1–2 page paper discussing:

Its use cases and benefits
Potential ethical issues (bias, consent, risk, explainability)
Your recommendations for building a responsible system

Include at least 2 references to current events or published research

Due: End of Week 10