TransCryber

Smart ears for tiny tear

Created on 23rd February 2025

•

Supports Single & First-Time Parents: Helps single and new parents understand their baby’s needs, reducing stress and making caregiving easier.
Assists Caregivers & Hearing-Impaired Parents: Babysitters, daycare staff, and deaf parents can rely on AI-driven insights to interpret baby cries accurately.
Detects Health Issues Early: Identifies cries related to pain, discomfort, or illness, allowing parents to seek medical help sooner.
Enables Remote & Smart Monitoring: Working parents can receive real-time alerts, and the tool can integrate with baby monitors for better care.
Reduces Stress & Improves Response Time: Eliminates guesswork, helping parents respond quickly and confidently, improving both baby’s comfort and parental mental health.

Limited Dataset & Audio Preprocessing
Finding a diverse labeled dataset for baby cries was tough. Plus, cleaning audio by removing background noise was time-consuming.
Training Speed & Compute Constraints
Training deep learning models in limited time and resources was challenging. We had to optimize the architecture and use transfer learning for faster results.
Real-Time Processing & Model Integration
Ensuring low-latency predictions while integrating the ML model with our Streamlit web app required debugging and optimization.
Accuracy in Classifying Baby Cries
Babies express mixed emotions while crying, making it difficult to classify cries correctly (hunger vs. pain). Balancing accuracy and speed was tricky.
UI/UX & Deployment Challenges
Designing an intuitive interface while deploying the model efficiently (locally or on a cloud server) within 36 hours was a major challenge.

Tracks Applied (1)

Our project aligns perfectly with the Streamlit track in MLH because it showcases how machine learning (ML) models can b...Read More

Major League Hacking

Technologies used

Python

Streamlit