🐟 Fish Disease Detection API

An AI-powered web application for detecting fish diseases from images using an InceptionV3 deep learning model.

✨ Features

• 🤖 AI-Powered Detection: Uses InceptionV3 model trained for fish disease classification
• ⚡ FastAPI Backend: High-performance REST API with automatic documentation
• 🎨 Modern Web Interface: Responsive, user-friendly interface with drag-and-drop support
• 📊 Real-time Analysis: Instant predictions with confidence scores
• 🔍 Non-Fish Detection: Intelligent filtering to reject non-fish images
• 💊 Multiple Disease Classes: Detects 7 different fish diseases

🏥 Supported Disease Classes

1. Bacterial Red disease
2. Bacterial diseases - Aeromoniasis
3. Bacterial gill disease
4. Fungal diseases Saprolegniasis
5. Healthy Fish
6. Parasitic diseases
7. Viral diseases White tail disease

📋 Prerequisites

• Python 3.8 or higher
• Model file: inceptionv3_fish_final.h5 (should be in /Users/riyadafromspace/Documents/)

🛠️ Installation

Quick Setup

# Navigate to project directory
cd /Users/riyadafromspace/Documents/api_integration

# Create virtual environment
python3 -m venv venv
source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

Manual Setup

# Create virtual environment
python3 -m venv venv
source venv/bin/activate  # On macOS/Linux
# or
venv\Scripts\activate  # On Windows

# Install required packages
pip install fastapi uvicorn tensorflow pillow python-multipart numpy

🚀 Running the Application

Start the Server

# Make sure virtual environment is activated
source venv/bin/activate

# Run the server
python main.py

Or using uvicorn directly:

uvicorn main:app --reload --host 0.0.0.0 --port 8000

Access the Application

• Web Interface: http://localhost:8000
• API Documentation: http://localhost:8000/docs
• Alternative Docs: http://localhost:8000/redoc

📡 API Usage

Health Check

curl http://localhost:8000/health

Response:

{
  "status": "healthy",
  "model_loaded": true,
  "num_classes": 7
}

Predict Disease

Using cURL:

curl -X POST "http://localhost:8000/predict" \
  -H "Content-Type: multipart/form-data" \
  -F "file=@/path/to/fish/image.jpg"

Using Python:

import requests

url = "http://localhost:8000/predict"
files = {'file': open('fish_image.jpg', 'rb')}
response = requests.post(url, files=files)
print(response.json())

Response:

{
  "success": true,
  "predicted_class": "Bacterial gill disease",
  "confidence": 0.8523,
  "confidence_percentage": "85.23%",
  "is_fish": true,
  "is_uncertain": false,
  "entropy": 0.42,
  "top_2_diff": 0.65,
  "image_std": 45.3,
  "all_predictions": [...],
  "message": "Bacterial gill infection detected..."
}

Get Available Classes

curl http://localhost:8000/classes

🎨 Web Interface Usage

1. Open http://localhost:8000 in your browser
2. Upload a fish image by clicking or dragging and dropping
3. Analyze the image by clicking the "Analyze Image" button
4. View detailed results including:
   • Disease prediction
   • Confidence score
   • Treatment recommendations
   • All class probabilities

🔧 Configuration

Model Path

By default, the model is loaded from:

/Users/riyadafromspace/Documents/inceptionv3_fish_final.h5

To change the model location, update main.py:

MODEL_PATH = os.getenv("MODEL_PATH", "/path/to/your/model.h5")

Non-Fish Detection Thresholds

Adjust detection sensitivity in main.py (around line 165):

is_non_fish = (
    (confidence < 0.40) or  # Minimum confidence threshold
    (confidence < 0.55 and normalized_entropy > 0.75) or
    (img_std < 15) or  # Image uniformity threshold
    (confidence < 0.50 and top_2_diff < 0.15)
)

📁 Project Structure

api_integration/
├── main.py                 # FastAPI application
├── requirements.txt        # Python dependencies
├── README.md              # This file
├── .env.example           # Environment variables template
├── .gitignore            # Git ignore rules
├── run.sh                # Quick start script (Unix)
├── setup.sh              # Setup script (Unix)
├── static/
│   ├── index.html        # Web interface
│   ├── style.css         # Styling
│   └── script.js         # Client-side logic
└── venv/                 # Virtual environment (created on setup)

🔍 How It Works

Non-Fish Image Detection

The system uses multiple indicators to detect non-fish images:

1. Confidence Score: Model's certainty (< 40% = likely not fish)
2. Entropy: Prediction uncertainty (high entropy = random/confused)
3. Image Uniformity: Pixel variance (too uniform = solid color/pattern)
4. Prediction Spread: Gap between top predictions (small gap = confused)

Image Preprocessing

1. Convert to RGB (if needed)
2. Resize to 224x224 pixels
3. Apply InceptionV3 preprocessing (scales to -1 to 1)
4. Add batch dimension for model input

Prediction Pipeline

1. Upload image via web interface or API
2. Validate file type
3. Preprocess image
4. Predict using InceptionV3 model
5. Validate if image is actually a fish
6. Return results with confidence scores and recommendations

🐛 Troubleshooting

Model Not Loading

Issue: Model not loaded error

Solutions:

• Check model file path in main.py
• Ensure model file exists: ls -lh /Users/riyadafromspace/Documents/inceptionv3_fish_final.h5
• Verify file permissions

Server Won't Start

Issue: Port already in use

Solutions:

# Find process using port 8000
lsof -ti:8000

# Kill the process
kill -9 $(lsof -ti:8000)

# Or use a different port
uvicorn main:app --port 8001

Predictions Always "Not a Fish"

Issue: Real fish images rejected

Solutions:

• Check image quality (blur, lighting, resolution)
• Review server logs for metrics
• Lower confidence threshold in main.py

Low Confidence Predictions

Issue: All predictions < 50%

Solutions:

• Use clearer, well-lit images
• Ensure fish is clearly visible
• Check if model was trained on similar images
• Verify preprocessing matches training

📝 Requirements

fastapi>=0.100.0
uvicorn[standard]>=0.23.0
tensorflow>=2.13.0
pillow>=10.0.0
python-multipart>=0.0.6
numpy>=1.24.0

🤝 Contributing

Feel free to submit issues and enhancement requests!

📄 License

This project is for educational and research purposes.

🔗 API Documentation

Once the server is running, visit:

• Swagger UI: http://localhost:8000/docs
• ReDoc: http://localhost:8000/redoc

Interactive API documentation with request/response examples.

💡 Tips

• Best Results: Use clear, well-lit photos of fish showing visible symptoms
• Image Quality: Higher resolution images (but will be resized to 224x224)
• File Formats: JPEG, PNG, JPG supported
• Non-Fish Images: System will automatically reject images that don't appear to be fish

🚀 Quick Start Script

For Unix/Linux/macOS:

chmod +x run.sh
./run.sh

This will activate the virtual environment and start the server.

---

Model: InceptionV3 (21.8M parameters, 130MB)
Framework: TensorFlow 2.x
API: FastAPI
Preprocessing: InceptionV3 standard preprocessing

For questions or issues, check the logs or API documentation.