Solving Medicine Identification: AI Medicine Strip Analyzer using YOLOv8, OCR & Gemini
Link to Live Project: https://medium.com/@surjeetlko5
Brief: An AI solution to scan medicine packaging, recognize the drug, and fetch real-time medical guidelines & safety warnings.
Introduction:
Illegible handwriting or hard-to-read text on medicine strips can lead to harmful dosing. I built an AI application to instantly analyze medicine strips from photo captures and retrieve easy-to-understand drug information.
How It Works:
1. Package Detection: Trained a custom YOLOv8 model to detect and isolate medicine strips/foils from various lighting and angles.
2. Text Extraction: OCR identifies brand names, drug compounds, and expiration dates on the packaging.
3. LLM Reasoning: Extracted text is fed to a FastAPI back-end that queries Gemini API to synthesize patient guidelines (uses, dosage, side effects).
4. Safety First: The pipeline generates explicit alerts about allergies and expiry checks.
Tech Stack:
YOLOv8, OpenCV, OCR (PaddleOCR), Gemini API, FastAPI, Python.
Key Results:
End-to-end drug identification and explanation in under 3 seconds with critical warnings highlighted.
