FocusLearn AI

Inspiration

The rapid shift to online learning has improved accessibility but introduced a major challenge: loss of student attention and engagement. In virtual classrooms, educators lack real-time visibility into whether students are focused or comprehending the material. This inspired us to build FocusLearn AI, a system that actively monitors attention and evaluates learning during live online sessions.

What it does

FocusLearn AI is an intelligent assistant for online learning that:

• Tracks student attention in real time using eye and facial analysis
  
• Detects distractions and yawning, and prompts learners to refocus
  
• Generates AI-based MCQ quizzes from live class content
  
• Provides instant assessment to measure understanding
  
• Integrates seamlessly with existing online meeting platforms
  

How we built it

• Frontend: Next.js with WebRTC for real-time video streaming
  
• Computer Vision: TensorFlow.js and MediaPipe for eye tracking and facial landmark detection
  
• AI Quiz Generation:
  • Keyword extraction using PKE
  • Sentence-keyword mapping with FlashText
  • Question generation using T5 Transformer
  • Distractor generation using Sense2Vec
• Backend: Flask with NLP libraries such as spaCy and NLTK
  
• Integration: Secure authentication and seamless classroom workflow
  

Challenges we ran into

• Maintaining real-time eye tracking performance without browser lag
  
• Reducing false positives in distraction and yawning detection
  
• Generating high-quality, non-repetitive MCQs
  
• Synchronizing live video, speech-to-text, and quiz generation
  
• Managing multiple AI pipelines under hackathon time constraints
  

Accomplishments that we're proud of

• Successfully implemented real-time attention tracking in live sessions
  
• Built a complete AI-driven quiz generation pipeline
  
• Created a non-intrusive and user-friendly monitoring experience
  
• Integrated computer vision and NLP into a single platform
  
• Delivered a working prototype within limited hackathon time
  

What we learned

• Real-time AI systems require balancing accuracy and performance
  
• Browser-based machine learning can be powerful when optimized
  
• Strong keyword–context alignment is critical for quality question generation
  
• Modular system design simplifies iteration and debugging
  
• Cross-functional collaboration is key in AI-driven projects
  

What's next for FocusLearn AI

• Personalized engagement analytics dashboards for educators
  
• Adaptive quiz difficulty based on learner performance
  
• Privacy-first, on-device inference optimization
  
• Integration with LMS platforms like Moodle and Google Classroom
  
• Long-term attention and engagement trend analysis
  

Built With

• flashtext
• getstream.io
• mediapipe
• natural-language-processing
• next.js
• nltk
• pke
• sense2vec
• t5
• tailwind-css
• typescript
• webrtc