cairo, egypt

Designed and implemented a real-time smile detection system for a Japanese elder care app using MediaPipe and custom deep learning techniques. The project addressed the challenge of detecting subtle facial expressions common among elderly Japanese individuals, where traditional OpenCV models fell short.

Key highlights:

Utilized MediaPipe Face Mesh for accurate facial landmark detection.

Engineered features based on mouth geometry to improve smile classification.

Trained a lightweight model tailored for real-time inference.

Achieved higher precision and responsiveness compared to standard models.

Integrated the solution into the client’s application with smooth performance on low-power devices.

The project enhanced user emotion tracking and supported improved emotional care for elderly users.