SafeBlurr | Devpost

climate change rally
BLM protest
Video processing demo

SafeBlurr: Anonymizing Video Recordings for a Safer Tomorrow

Presentation

https://docs.google.com/presentation/d/15wHEWzas6ybikhepC_IpRKPab1zvoA0ucC0HOyyDb9A/edit?usp=sharing

Introduction

Video is a powerful medium for social change. However, from women’s rights rallies to BLM protests, our current technologies do little to protect those involved. We introduce SafeBlurr as a groundbreaking app that preserves privacy and enhances user experience in video content.

What is SafeBlurr?

SafeBlurr is a cutting-edge app designed to anonymize video recordings, essential for journalists, activists, and anyone involved in public filming. It maintains the narrative's impact while safeguarding individual identities.

Features

Advanced Anonymization: using computer vision powers SafeBlurr’s face detection and automatic filter application, also automatically distorts voice audio
Customizable AI Speech-to-Text Subtitle Generation: For accessible transcribility and interview production value.
User-Friendly Mobile Interface allows users to easily take videos and process them like a social media app.

Challenges we ran into:

Finding the right bounds for facial blurring: Often the blur would flicker or be hazy around edges, so we needed to carefully trial and error the amount of error and blur-buffering before and after each AI recognition
Connecting Flask and React Native: Sending and processing videos across HTTP is resource consuming and unfamiliar, so we struggled to find the right protocol and efficient loading
Voice pitching change: To anonymize voices, we had to change pitch, which meant speed would change. So we needed alternative/secondary corrective measures.
Syncing Video and Audio: OpenCV doesn't support audio, only visual edits. So we needed to use extra libraries and workarounds to sync back our edited audio and video after processing.

Target Audience

Journalists and Media Professionals: Ensuring source and subject anonymity.
Activists and Protesters: Documenting events while protecting participant identities.
Public Servants and Law Enforcement: For responsible data handling.
General Public: For ethical recording practices in public spaces.

Built With:

React-Native
Flask
OpenCV, MediaPipe
Google Cloud Storage, Vision API

Future plans:

More options for customizing the degree of anonymization: censoring names, level of blurring, etc
Better security and privacy: More focus on encrypting the video transfer
Make the software needed to anonymize local to the app, to not risk network interceptions
Faster load times and response times for video processing

SafeBlurr is not just an app; it's a commitment to ethical media sharing. It's the perfect tool for ensuring truth and integrity in media, while upholding the utmost standards of privacy and safety.