PharmaScan

PharmaScan is a Streamlit-based web application that uses a custom YOLOv5 model to detect and extract handwritten medication names and their details from scanned medical prescriptions. It provides a simple, user-friendly interface to upload prescription images, analyze them with AI, and display the detected medications along with their confidence scores.

---

Table of Contents

• Features
• Screenshots
• How It Works
• Installation and Usage
• Model and Dataset
• Project Structure
• Contact
• License

---

Features

1. Image Upload
   Users can upload prescription images directly from their local system.

2. Real-Time Detection
   A custom YOLOv5 model processes the uploaded image, detecting:

   • Drug Names
   • Quantities
   • Dosage Information

3. Interactive Interface

   • View the original image and the detection results side-by-side.
   • A table listing detected medications, confidence scores, and other relevant info.

4. Streamlit Integration

   • Easy to run locally or deploy on popular cloud platforms (Streamlit Sharing, Heroku, etc.).
   • Responsive web UI.

---

Screenshots

Below are the main views of PharmaScan, displayed one by one with brief descriptions:

PharmaScan Home

The main landing page showing the title, introduction, and buttons for scanning or choosing an image.

Sample Prescription

A prescription image before detection, demonstrating how text appears in the original scan.

Result Overview

Side-by-side comparison of the original prescription and the detected bounding boxes, along with the list of detected medications.

Detection Result

The final processed image showcasing bounding boxes around identified drug names and other relevant information.

---

How It Works

1. Upload: The user selects or drags and drops a prescription image onto the web app.
2. Processing: The YOLOv5 model runs object detection on the image to locate and label handwritten medication details.
3. Display: PharmaScan displays the original image alongside the detection output. A table below shows each detected medication and confidence score.

---

Installation and Usage

1. Clone the Repository

   git clone https://github.com/mohamed-ladjal-AI/PharmaScan.git
   cd PharmaScan

2. Create and Activate a Virtual Environment (Optional but Recommended)

   # For example, using conda:
   conda create -n pharmascan python=3.9
   conda activate pharmascan

3. Install Dependencies

   pip install -r requirements.txt

4. Run the Streamlit App

   streamlit run web-app.py

   The app will open in your default browser at http://localhost:8501 (or another port if 8501 is in use).

---

Model and Dataset

• Model: This project uses a custom YOLOv5 model fine-tuned for prescription medication name detection.
• Dataset: The dataset used to train the YOLOv5 model is private and cannot be published publicly.
  If you urgently need access to the dataset for research or collaboration, please contact me.

---

Project Structure

PharmaScan/
├── assets/
│   ├── screencapture1.png
│   ├── screencapture2.png
│   ├── screencapture3.png
│   ├── result.jpg
│   └── logo.png
├── yolov5/
│   └── ... (YOLOv5 code and files)
├── web-app.py
├── requirements.txt
├── README.md
└── ...

• web-app.py: Main Streamlit application file.
• yolov5/: YOLOv5 code and scripts for detection.
• assets/: Contains images for the README and the application logo.

---

Contact

• Author: Mohamed Ladjal
• Email: [email protected]

Feel free to reach out if you have questions about the model, the dataset, or potential collaboration opportunities.

---

License

This project is licensed under the MIT License.
You’re free to use, modify, and distribute the code as long as the original license is included.