Build a Face-Anonymization Tool for a Civic-Tech Newsroom
Overview
What this challenge is about.
Use a pretrained face detector (RetinaFace or YOLOv8-face is fine). Build a Python tool with a Gradio or Streamlit UI that: (1) detects faces in an uploaded photo, (2) shows detections to the journalist for accept/reject, (3) outputs the blurred image + a JSON manifest of which detections were applied. Evaluate detection precision/recall on a small held-out set of 100 newsroom-style photos. Deliver the tool, evaluation notebook, and a 3-page editorial-team rollout doc.
The Brief
What you'll do, and what you'll demonstrate.
Ship a reviewer-in-the-loop face-anonymization tool that cuts hand-blurring time by 75 percent without missing high-risk faces.
Earning criteria — what you'll demonstrate
- Integrate a pretrained face detector into a working tool
- Build a reviewer-in-the-loop UI that respects journalist workflow
- Evaluate detection with precision/recall on a representative test set
- Document a privacy-sensitive tool for non-technical users
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
Careers
Roles this prepares you for.
Real titles. Real skill bridges. Pick the one closest to your trajectory.
AI Engineer
Wrapping a pretrained CV model in a reviewer-friendly tool is the day-one AI-engineering work at any media, legal, or compliance-AI company.
This challenge sharpens
- object-detection
- image-processing
- ui-prototyping
Computer Vision Engineer
Evaluating detection precision/recall on a domain-specific set is the CV-engineer skill that distinguishes shippable from demo.
This challenge sharpens
- object-detection
- model-evaluation
- opencv
AI Product Designer
Designing a reviewer UI that respects journalist workflow is exactly the AI product designer's craft of fitting AI into a human process.
This challenge sharpens
- ui-prototyping
- image-processing
- object-detection