Dragon Fruit Ripeness Detection System

---

Overview

The Dragon Fruit Ripeness Detection System is a production-ready AI system that predicts the ripeness level of dragon fruits using deep learning.

It combines:

• Deep Learning (EfficientNet)
• FastAPI backend for inference
• Docker containerization for deployment
• Intelligent decision layer for actionable insights

---

Features

• Image-based ripeness detection

• Ripeness percentage prediction

• Intelligent output layer:

• Quality scoring (0–100)

• Market grading (A+, A, B+, B, C)

• Shelf-life estimation

• Price categorization

• Harvest recommendation

• Out-of-Distribution (OOD) detection

• Optimized CPU inference

• REST API for mobile/web integration

• Fully Dockerized for portability

---

System Architecture

ML Model (EfficientNet) ↓ FastAPI Backend ↓ Inference + Intelligence Layer ↓ Docker Container ↓ Cloud Deployment (Render) ↓ Mobile / Client Integration

---

Tech Stack

• ML Framework: PyTorch, timm
• Backend: FastAPI
• Deployment: Docker
• Image Processing: PIL
• Numerical Ops: NumPy

---

Getting Started

1. Clone Repository

git clone https://github.com/your-username/dragon-fruit-ai.git
cd dragon-fruit-ai/backend

---

2. Run with Docker

docker build -t dragon-fruit-api .
docker run -p 8000:8000 dragon-fruit-api

---

3. Access API

http://localhost:8000/docs

---

API Usage

Endpoint

POST /predict

---

Request

• Type: multipart/form-data
• Field: file (image)

---

Example Response

{
  "ripeness_percent": 78,
  "grade": "A",
  "quality_score": 82,
  "price_category": "Premium",
  "shelf_life_days": 3,
  "harvest_recommendation": "Ready"
}

---

Error Handling

Scenario	Response
Invalid image	400
Internal error	500
OOD input	{ "status": "invalid" }

---

Testing

• Swagger UI available at /docs
• Supports real-time image upload
• Validated with real dataset samples

---

Project Structure

dragon-fruit-ai/
├── backend/
│   ├── app/
│   ├── models/
│   │   └── ripeness_model.pth
│   ├── Dockerfile
│   ├── requirements.txt
│   └── .dockerignore
├── ai-engine/

---

Deployment

The system is designed for deployment on cloud platforms like Render.

Key Features:

• Dynamic port handling
• Docker-based deployment
• Production-ready architecture

---

Contributing

We welcome contributions from developers, ML engineers, and students!

Steps:

1. Fork the repository
2. Create a new branch
3. Make your changes
4. Submit a Pull Request

---

Open Issues

Check out:

• good-first-issue for beginners
• help wanted for contributors
• ml, backend, devops for domain-specific tasks

---

Future Improvements

• API authentication & rate limiting
• Mobile app integration
• Batch inference support
• Model explainability (Grad-CAM)
• Multi-fruit detection system
• CI/CD pipeline automation

---

Learnings & Challenges

This project involved solving real-world engineering problems:

• Model architecture mismatch (timm vs torchvision)
• Docker dependency conflicts
• Large image optimization (CUDA → CPU)
• Network timeout handling
• Cloud deployment constraints

---

Highlights

• End-to-end ML system (training → deployment)
• Production-ready backend
• Dockerized and optimized
• Real-world problem solving

---

Contact

For collaboration or queries:

• GitHub: https://github.com/defender-777

---

Support

If you find this project useful:

Star ⭐ the repository Share with others Contribute 🚀

---