About PoultryDetect

The PoultryDetect project is an innovative deep learning-based system designed to classify common poultry diseases: Salmonella, New Castle Disease, Coccidiosis, and Healthy. Our core objective is to empower farmers and veterinary professionals with a rapid, accurate, and accessible diagnostic tool that leverages advanced computer vision. By simply uploading an image of affected poultry, users can receive an immediate diagnosis, enabling swift action to manage flock health and prevent widespread outbreaks.

This system significantly reduces the time and specialized expertise traditionally required for disease identification, ensuring healthier poultry, improved productivity, and reduced economic losses for farms of all sizes, from rural communities to large commercial operations.

Key Features

Core functionalities that make PoultryDetect powerful and efficient.

Accurate Disease Classification

Utilizes advanced transfer learning models (VGG16, VGG19, ResNet50) for high-precision diagnosis.

User-Friendly Web Interface

A simple and intuitive Flask-based web application allows for easy image uploads and instant result display.

Rapid Diagnostics

Provides near real-time predictions, enabling prompt intervention and disease control.

Accessible Tool

Designed to be accessible to farmers in various settings, including those with limited access to traditional veterinary services.

Scalable Architecture

Built on open-source frameworks, allowing for future expansion and integration with more complex systems.

LLM Integration (Gemini API)

Provides AI-generated disease information and quick action plans for immediate insights.

Technology Stack

This project is built on a foundation of robust, open-source technologies, ensuring a scalable and maintainable solution.

Python 3.x Flask TensorFlow Keras VGG16/VGG19/ResNet50 NumPy Pandas Pillow OpenCV HTML5 CSS3 (Tailwind CSS) JavaScript Gemini API (LLM)

Training Dataset Composition

Distribution of images across the four poultry health categories.

How It Works

From setup to execution, here's a look under the hood.

Setup Instructions

1. Install Anaconda Navigator: Download and install Anaconda Navigator.

2. Install Python Packages: Open Anaconda Prompt (as administrator) and install the necessary libraries:

pip install numpy pandas scikit-learn matplotlib scipy seaborn tensorflow Flask keras-tuner opencv-python Pillow

3. Dataset Acquisition: Obtain the poultry disease dataset (e.g., from Kaggle).

4. Model Training: Train your deep learning model and save it as healthy_vs_rotten.h5.

Team Details

Team ID: LTVIP2025TMID41509
Team Size: 3
Team Leader: Vaddimeyani Ajay Kumar
Team Member: Vadlamudi Bavanchandu
Team Member: Y Lavanya

Project Directory Structure

Project_folder/
├── static/
│   ├── assets/       # General static assets (e.g., logos, shared images)
│   ├── forms/        # Form-related static files (if any)
│   └── uploads/      # Directory to store uploaded images for prediction
├── templates/
│   ├── about.html    # This page: Project overview, features, etc.
│   ├── contact.html  # The image upload and prediction display page
│   └── index.html    # The main landing page
├── app.py            # Main Flask application script
├── healthy_vs_rotten.h5 # The trained machine learning model
└── Readme.txt        # Project documentation and setup instructions

Running the Application

1. Navigate to Directory:

cd path/to/your/Project_folder

2. Run the Application:

python app.py

3. Access Web Page: Open a web browser and go to http://127.0.0.1:5000.

Core Prediction Logic

Image Upload: A farmer accesses the PoultryDetect web application and uploads an image of poultry (e.g., a sick bird, a sample).
Server-Side Processing: The Flask backend receives the uploaded image and securely saves it.
Image Preprocessing: The image is automatically resized, converted into a numerical array, and normalized to match the input requirements of the deep learning model.
Model Inference: The preprocessed image data is fed into the trained transfer learning model (e.g., VGG16). The model analyzes the visual features within the image.
Prediction: The model outputs a probability distribution across the four disease categories. The highest probability determines the predicted disease.
Result Display: The predicted disease label (e.g., "Coccidiosis", "Healthy") is sent back to the frontend and displayed to the farmer, along with the uploaded image for confirmation.

Future Vision

Potential enhancements to expand the project's capabilities.

➔

Expanded Disease Classification:

To include a wider range of poultry diseases and health conditions.

➔

Mobile Application Development:

Creating a native mobile app for easier on-the-go diagnosis directly from the farm.

➔

Prediction History & Analytics:

Implementing a database to store past predictions, allowing farmers to track flock health over time and identify patterns.

➔

Integration with IoT:

Connecting with environmental sensors to provide a holistic view of poultry health, factoring in conditions like temperature and humidity.

➔

Veterinary Consultation Integration:

Potentially linking with veterinary telemedicine platforms for expert consultation based on AI diagnosis.

➔

User Authentication:

Implementing secure user accounts for personalized data management and privacy.