Inspiration
Prisoners of war (POWs) have historically used subtle cues, like blinking Morse code, to send distress messages during captivity. In high-stakes environments like interrogations, identifying emotional stress and subliminal messaging can mean the difference between continued suffering and rescue. We wanted to create a system that empowers intelligence agencies and humanitarian efforts to identify these hidden signals using AI, computer vision, and emotion detection, bringing technology to the service of justice and freedom.
What it does
Subliminity is an AI-powered tool that analyzes interrogation footage in two ways:
- Live Emotion Recognition: Uses real-time facial analysis to detect stress, fear, or discomfort—indicators of coercion or deception.
- Video Morse Code Detection: Analyzes eye movement and blinking patterns in recorded videos to detect subliminal Morse code messages potentially used by POWs to signal distress.
The system presents a user-friendly web interface with two modules, enabling seamless analysis of both live and recorded footage.
How we built it
- Flask (Python) for backend routing and API endpoints.
- DeepFace for real-time emotion classification from webcam feeds.
- MediaPipe for precise eye and facial landmark tracking.
- OpenAI API for natural language support, contextual analysis, and system decision reinforcement.
- JavaScript & HTML/CSS to build an intuitive front-end tabbed interface.
- Custom Algorithms for parsing blink timings into Morse code and distinguishing between dots and dashes using temporal thresholds.
I also employed threading and OpenCV to ensure efficient frame capture and smooth real-time analysis.
Challenges we ran into
- Interpreting Blink Data: Differentiating between natural blinking and intentional Morse sequences required fine-tuning timing thresholds and filtering noise.
- Emotion Detection Lag: Real-time performance with DeepFace was initially slow, so we had to optimize frame sampling rates.
- Front-end/Back-end Sync: Ensuring seamless video capture from the browser to the backend, while maintaining analysis accuracy, was a cross-domain challenge.
- Ethical Safeguards: Designing a tool with such sensitivity demanded attention to false positives and user privacy.
Accomplishments that we're proud of
- Built a fully functional live and recorded analysis system in one interface.
- Developed a Morse code detection pipeline from eye tracking data—rare and highly specialized.
- Created a tool with real-world impact for defense and humanitarian efforts.
What we learned
- The power of combining emotion AI and optical tracking to reveal hidden information in video data.
- How to convert temporal patterns into semantic messages (e.g., blinks to Morse code).
- The importance of designing with ethical considerations, especially in systems related to surveillance and personal safety.
- How to better use Flask, DeepFace, MediaPipe, and OpenAI in a cohesive pipeline.
What's next for Subliminity
- Potential automated alert systems when Morse code is detected with high confidence.
- Additional fully powered machine-learning models for other declarations of help.
GitHub Copliot was used in this project.
Log in or sign up for Devpost to join the conversation.