StreamSafe: A Privacy Protection Tool for Live Streams

Inspiration 🌟

The idea for StreamSafe was inspired by the frustration and concern we felt after watching our favorite streamers face harassment when their real locations were exposed during live streams. This risk of privacy compromise made us realize that there was a need for a solution that could automatically blur sensitive content in real-time, without affecting the stream's quality. The blurring that’s often used by streamers today is usually jarring and ineffective, leading to a poor viewer experience. Thus, we set out to create a tool that would offer seamless, effective privacy protection, keeping streamers safe while maintaining a high-quality streaming experience.

What It Does 🚀

StreamSafe is a real-time privacy protection tool for live streamers. It automatically detects and blurs sensitive content in video streams, including:

License Plates 🚗
Street Signs 📍
Building Numbers 🏠

The tool uses cutting-edge AI powered by YOLO for real-time object detection and OpenCV for video frame processing. StreamSafe works seamlessly with popular platforms like Twitch, YouTube, and Kick, ensuring streamers’ privacy is protected while maintaining high video quality.

Key Features:

Live Streaming: Real-time privacy protection during live streams.
WebRTC Integration: Video streaming with WebRTC support.
Live Metrics: Real-time FPS, latency, and detection count monitoring.

How We Built It 💻

We built StreamSafe using the following technologies:

Streamlit for the frontend interface to provide an easy-to-use dashboard for streamers.
OpenCV for real-time video frame processing to blur sensitive content.
PyTorch & YOLO for real-time object detection to automatically detect and blur sensitive content in the live stream.
DroidCam enables live camera footage from phones streaming directly into the Streamlit app

Challenges We Ran Into ⚠️

We encountered several challenges during development, including:

Real-time video processing: Ensuring privacy protection was applied while maintaining video quality was one of the toughest challenges.
Integration: Combining WebRTC, YOLO, OpenCV, and Streamlit posed significant technical challenges, especially with handling different video formats and streaming protocols.
User Experience: Designing a user-friendly interface that allowed streamers to easily toggle privacy settings without interrupting the stream was tricky. We had to ensure the tool was intuitive and didn’t interfere with the streamer's workflow.

Accomplishments That We're Proud Of 🏆

Seamless real-time video processing: Achieving real-time object detection and blurring of sensitive content in live streams.
User-friendly design: Building a straightforward and effective user interface with Streamlit that allowed streamers to easily adjust privacy settings.

What We Learned 📚

AI and computer vision: We deepened our understanding of object detection and real-time video processing using PyTorch and YOLO.
Streamlit: Learned how to leverage Streamlit to build a user-friendly interface with real-time data updates.
Privacy in live streaming: Gained insights into the growing concern of privacy issues in live streaming and how effective privacy protection can enhance streamers' safety.

What's Next for StreamSafe 🚀

GPU Hardware Acceleration: Implement GPU-based hardware acceleration to reduce lag time and significantly improve the framerate, ensuring smoother and faster processing.
Expanded Content Detection: Add detection for more landmarks like McDonald's, Starbucks, and other popular global locations to further enhance privacy protection.
International Street Sign Recognition: Improve detection models by training them with street sign datasets from various countries. The current implementation is based on Singapore's distinctive green and blue street signs, and expanding this to include other countries will enhance global applicability.

With StreamSafe, we aim to create a safer and more privacy-conscious environment for streamers worldwide, ensuring they can focus on their content while we handle their privacy protection seamlessly.