Yolocilin v1.0.0

Yolocilin v1.0.0

First public release of Yolocilin — Medicine Box Detection, Identification and Information System.

Photograph a box → detect it (YOLOv8) → read the print (EasyOCR) or a barcode → match against a 1163-row TİTCK-enriched catalog → show the result on Android, with an optional short Gemini explanation.

Not medical advice. Always confirm with the official leaflet or a pharmacist.

Highlights

• Pipeline: YOLOv8n box detection → crop → OpenCV → EasyOCR (fast / accurate) → RapidFuzz → SQLite
• Matching policy: prefer not_found over a wrong drug (suffix OCR such as fen / alm is rejected)
• Barcode path: optional EAN-13 / DataMatrix lookup before OCR; Flutter live + photo scanner
• API: FastAPI — analyze, medicines, barcode, explain, scans
• Mobile: Flutter Android app (com.yolocilin.app) — gallery, camera, history, bilingual UI
• Ops: Docker Compose, GitHub Actions CI, production CORS / /docs hardening, rate limits
• Data: Kaggle dataset (CC BY 4.0, 395 images)

Versions

Surface	Version
API (/health)	1.0.0
Flutter / Android	1.0.0+1
Catalog	1163 medicines

Run

git clone https://github.com/Melikeda/yolocilin.git
cd yolocilin
git checkout v1.0.0

Python 3.11+, pip install -r requirements.txt, then python run_api.py.
Trained weights (best.pt) are not in Git — see models/README.md. Dataset images stay on Kaggle.

Full notes: CHANGELOG.md · Report 29

Out of scope

PostgreSQL, cloud HTTPS hosting, per-user scan auth, and iOS remain post-MVP.