MedScript Prescription OCR is an AI-powered application that extracts handwritten medicine names from prescription images and matches them against a cleaned drug dataset.
This project uses OCR and fuzzy matching to identify likely drug names from prescription images line by line.
- Upload prescription images
- Line-by-line OCR extraction
- OCR text cleaning
- Drug-name matching against a cleaned dataset
- Top matching results with confidence scores
- Python
- Gradio
- Hugging Face Transformers
- RapidFuzz
- OpenCV
- Pandas
This project uses a cleaned drug dataset hosted on Hugging Face.
Dataset repo:
hashan-7/medscript-drug-dataset
Dataset file:
data/master_drug_dataset.csv
- Upload a prescription image
- Segment the image into text lines
- Run OCR on each line
- Clean extracted text
- Match likely drug names
- Display top results
Install dependencies:
pip install -r requirements.txt