TECHNOLOGY
A self-sufficient data pipeline linking clinic and AI
We run the closed loop of device → consented collection → ophthalmologist labeling → model training → device redeployment, all inside one company.
Technical differentiation
Hybrid vision AI
Fine-tuned MediaPipe Face Mesh·Iris pretrained models, EfficientNet multi-class classification of 5 appearance patterns, combined with a classical computer-vision baseline.
Standardized capture
Light-blocking, standardized-light devices block external light (≥95%) to secure reproducible data.
Anterior-segment eye age score
Age estimation based on the anterior segment rather than the fundus — leading a globally untapped area.
Self-sufficient data pipeline
An ophthalmologist labels directly, achieving data self-sufficiency without external medical-institution collaboration or outsourced labeling.
Core R&D quantitative targets
≥ 90%
Capture-quality auto-grading
≥ 0.80
Appearance-pattern macro-F1
≥ 90%
Eye & facial motion recognition
≥ 70%
Eye-age classification (per decade)
≥ 95%
External-light blocking uniformity
≥ 4.0 / 5.0
User satisfaction (UEQ-S)
Security · Infrastructure
ICatcher runs in the AWS Seoul region with KMS encryption, VPC isolation, and domestic-region storage, complying with the Korean Personal Information Protection Act.