Overview
RONA is an AI-powered on-device acne recovery tracking assistant for iOS. It helps users independently and objectively evaluate their skin recovery journey through multi-angle face scans, zone-based skin scoring, and side-by-side historical photo comparisons, all processed 100% locally on-device to ensure user privacy.
Technical Highlights
- Custom Dataset: Curated and validated a clinical dataset of 14,630 images with 101,730 bounding box annotations across 6 clinical acne classes (papule, pustule, blackhead, nodule, cyst, whitehead).
- Model Architecture: Trained YOLO26s (9.95M parameters, 22.8 GFLOPs) at 640×640 resolution.
- Edge AI / CoreML Export: Converted PyTorch weights to CoreML with INT8 quantization, yielding a compact 9.3 MB model for offline inference.
- iOS Architecture: MVVM + Feature-Based Modular Architecture, Swift 6, SwiftUI.
Metrics
| Metric | Value |
|---|---|
| Precision | 71.86% |
| Recall | 64.29% |
| mAP50 | 67.21% |
| mAP50-95 | 38.09% |
Last updated on September 16, 2026 at 12:23 PM UTC+7. See Changelog