Overview
What this challenge is about.
You receive 500 internal benchmark images (already cleared for use), each labelled with bounding boxes around face/tattoo/jewelry regions. Build a pipeline that detects these regions (a pre-trained face detector plus a small fine-tuned model for tattoos/jewelry), applies irreversible removal (Gaussian blur + content-aware fill is fine — do not return original pixels), and routes any low-confidence detections to a Streamlit manual-review tool. Report detection recall (must hit 99% per category), and a small irreversibility check (verify removal cannot be inverted by a sharpening filter). Deliver: pipeline, review tool, evaluation report, and a 2-page process doc.
The Brief
What you'll do, and what you'll demonstrate.
Make external image sharing safe and fast by automating irreversible de-identification with a clean human-review fallback.
Earning criteria — what you'll demonstrate
- Combine pre-trained and fine-tuned detectors in a privacy pipeline
- Reason about irreversibility as a property of image transforms
- Build a human-in-the-loop fallback for low-confidence detections
- Author a process doc that maps technical choices to compliance requirements
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 Safety Researcher
Owning a privacy-preserving pipeline with documented irreversibility checks is exactly the work AI safety researchers do at pharma, healthtech, and any regulated-data org.
This challenge sharpens
- image-de-identification
- privacy-preserving-vision
- evaluation
Computer Vision Engineer
Combining pre-trained and fine-tuned detectors plus a human-in-the-loop tool is core CV-engineer work at any vertical-vision vendor.
This challenge sharpens
- object-detection
- image-de-identification
- human-in-the-loop
Machine Learning Engineer
Building a confidence-routed inference pipeline with a manual fallback is the MLE skillset for any high-stakes ML deployment.
This challenge sharpens
- object-detection
- human-in-the-loop
- evaluation