Project Thread🏆 Great Malaysia AI Hackathon 2025
LinguaDaily — AI-powered immersive language learning platform with computer vision
🐙 View GitHubOverview
Comprehensive language learning platform combining computer vision, speech recognition, and real-time translation.
Tech Stack
FlaskAWS RekognitionAzure Speech Cognitive ServicesOpenCVn8n
Features
- Real-time object detection with AWS Rekognition
- Multi-camera support (laptop, phone IP webcam)
- 4-metric pronunciation scoring (Accuracy, Fluency, Completeness, Prosody)
- Adaptive curriculum for 13+ languages
Key Decisions
01
Chose Azure Speech over Google STT due to Azure's granular "Prosody" and "Fluency" metrics, which are essential for Syllable-level pronunciation feedback.
02
Utilized n8n to orchestrate the translation pipeline, allowing for rapid deployment of 13+ language localizations without hardcoding.
Learnings
→Mastered the integration of low-latency WebRTC streams with cloud-based computer vision endpoints.
→Learned to handle "Visual Noise" by implementing a confidence threshold in OpenCV to filter out background objects that aren't relevant to the lesson.