Introduction

Welcome to our Plant Classification Web Application aka AushadHub! This app uses a pre-trained deep learning model to classify images of plants.

AushadHub Overview 🌿 :

AushadHub is a project designed to address the challenge of identifying medicinal herbs. It utilizes a Machine Learning Model based on ResNet, achieving a validation accuracy of 98% and a testing accuracy of 96%. The dataset used for training and testing comprises over 1500 images spanning 30 different medicinal herb species.

Description:

FrontEnd:

The front end of AushadHub is developed using Flask, HTML, CSS, and a touch of vanilla JavaScript.

Backend:

The backend consists of an end-to-end Machine Learning solution dedicated to solving the problem of medicinal herb identification. It employs ResNet trained on a custom dataset, achieving a validation accuracy of 98%. The dataset used for training and testing consists of over 1500 images representing 30 different medicinal herb species.

📁 Project Structure

• Hackmol5/
  • app.py: Flask application script
  • model/
    • resnet50-transfer.pth: Pre-trained ResNet50 model checkpoint
  • static/
    • upload/: Folder for user-uploaded images
  • templates/
    • index.html: Homepage template
    • result.html: Result page template
  • info1.csv: CSV file containing class labels and descriptions
• requirements.txt: List of Python dependencies

🚀 Quick Start

1. Clone the repository:

git clone [https://github.com/abhinav0git/HackMol5.git](https://github.com/abhinav0git/HackMol5)
cd HackMol5

2. Install dependencies:

pip install -r requirements.txt

3. Run the app:

python app.py

🤖 Model Details

The app uses a pre-trained ResNet50 model for plant classification. Model checkpoint: Hackmol5/model/resnet50-transfer.pth

Tech Stack Used:

• Python 3
• Flask
• Pillow
• HTML
• CSS
• JavaScript
• GitHub

📝 Additional Information

• Class labels and descriptions are stored in the info1.csv file.
• Ensure to create a Python environment tailored to your machine specifications, especially if you are using VSCode.