CNN Image Classification Web Application

This project is a web application that classifies potato leaf health based on uploaded images. It can detect Potato Early Blight, Potato Late Blight, and Healthy Potato Leaves, providing a confidence score for each prediction.

Project Overview

• Frontend: Built with React, featuring a user-friendly interface for image upload via drag-and-drop or file selection.
• Backend: FastAPI serves as the API layer, facilitating communication with the TensorFlow Serving model for real-time predictions.
• Model Deployment: A pre-trained CNN model, hosted on TensorFlow Serving, processes input images and returns classification results.

Dataset

The model is trained on the PlantVillage Dataset, which includes labeled images of potato leaves with various conditions, helping to differentiate diseased and healthy samples. p.s you can also find the Dataset here as a ZIP file: C:<your directory>\potato_classifier\training\PlantVillage.zip

Application Workflow

1. Model Training: The CNN model was trained using the PlantVillage dataset and saved for deployment.
2. Model Deployment: The trained model is deployed on TensorFlow Serving, accessed through FastAPI to serve predictions to the frontend.
3. Web Interface: Users upload an image of a potato leaf, and the app displays the classification and confidence score.

Getting Started

To get a local copy of the project running, follow these steps:

Prerequisites

1. Clone the Repository:

   git clone https://github.com/BEKtesfish/CNN-Image-Classification-Web-Application.git
   CNN-Image-Classification-Web-Application

2. Create and Activate a Virtual Environment:

   python -m venv venv
   source venv/bin/activate  # On Windows use `venv\Scripts\activate`

3. Install Backend Dependencies:

   pip install -r requirements.txt

Running the Application

1. Setup TensorFlow Serving: To serve the model with TensorFlow Serving, use Docker:

    docker run -t --rm -p 8501:8501 -v C:<your repository>\potato_classifier\models:/models tensorflow/serving --rest_api_port=8501 --model_config_file=/potato_classifier/models.config

   For more details, see the TensorFlow Serving Documentation.

2. Run the FastAPI Backend: Inside the api directory, start the backend server:

   cd api
   python main-tf-serving.py

3. Run the Frontend: Navigate to the frontend directory, install dependencies, and start the React app:

   cd ../front
   npm install
   npm start

4. Access the Application: Open your browser and go to http://localhost:3000. Here, you’ll find a drag-and-drop section where you can upload an image of a potato leaf to receive predictions and a confidence score

Additional Resources

Model Training Notebook: In the training folder,contains the Jupyter notebook used for training the CNN model. and the 1st version of the trained model can be found in the models folder