Alibi

The Problem

Anyone publishing video content with unidentified individuals has a clear legal obligation under GDPR: faces must be obscured before publication. The problem is that doing it manually is unsustainably slow. A 10-minute video with 20 people on screen requires hours of frame-by-frame work. For journalistic content, surveillance footage, or field research, this becomes a real bottleneck.

The risk is not just time: it's easy to miss a face in a crowded scene or a blurred frame. A single omission in a legal or journalistic context can have serious consequences.

The Solution

I built a system based on convolutional neural networks (CNN) that automatically detects every face in the video and applies blur frame by frame. Detection doesn't stop at the first frame: it tracks each face throughout the video's duration, even as the person moves or changes angle.

The most critical part was reducing false positives. Not every face-shaped object is a face: posters, reflections, mannequins, and masks are often detected as false positives by standard models. I implemented a multi-threshold scoring system that evaluates each detection based on confidence, size, proportions, and temporal consistency across frames. Result: false positives reduced by 80% compared to baseline models.

Batch processing handles entire folders of videos and images in queue. For each file the output is a blurred version ready for publication, with a log of detected faces and uncertainty areas for manual review.

The Result

A 10-minute video with dozens of people is processed in a few minutes instead of hours. The system is used for journalistic content with anonymous sources, surveillance footage shared with authorities, and university research with sensitive visual data. Intelligent validation minimizes manual intervention while keeping the ability to review every flagged frame.