☕ Coffee Grind AI Analyzer

An industrial-grade Deep Learning solution for Coffee Particle Size Distribution (PSD) analysis. This project transforms a traditional OpenCV-based tool into a modern, containerized AI service capable of high-precision grind size measurement using YOLOv8 segmentation.

---

🚀 Highlights

• AI Inference Engine: Integrates YOLOv8-segmentation for precise particle detection and size calculation.
• Microservices Architecture: Separate containers for API (FastAPI), Task Queue (Celery), and In-memory Store (Redis).
• Asynchronous Processing: Handles heavy AI inference tasks in the background to ensure a smooth UI experience.
• Manual Calibration Tool: Built-in canvas-based scaling tool in React to convert pixels to real-world millimeters (mm).
• Dockerized Environment: Fully containerized deployment with optimized CPU-only PyTorch builds.

---

🛠️ Technical Architecture

Note

The diagram above shows the full microservices flow. If your viewer doesn't render the Mermaid code below, please refer to the image.

graph LR
    A[React Frontend] -->|Upload + Calibration| B[FastAPI API]
    B -->|Dispatch Task| C[Redis Queue]
    C -->|Fetch Task| D[Celery AI Worker]
    D -->|YOLOv8 Prediction| D
    D -->|Calc PSD| D
    D -->|Store Result| C
    B ---|Poll Result| C

Loading

---

📦 Project Structure

• api/: Backend core logic.
  • main.py: RESTful API endpoints.
  • worker.py: AI inference and PSD calculation logic.
• frontend/: Modern React + Vite + Tailwind CSS dashboard.
• synthetic_data/: Scripts and utilities for generating AI training data.
• Dockerfile & docker-compose.yml: Container orchestration.

---

⚙️ Quick Start (Docker)

Ensure you have Docker Desktop installed.

1. Clone and Build:

   docker-compose up --build -d

2. Access the Services:

   • Frontend: http://localhost:5173
   • API Docs (Swagger): http://localhost:8000/docs

---

🎨 Frontend Setup (Manual)

If you wish to run or develop the frontend separately without Docker:

1. Prerequisites: Node.js installed.
2. Install Dependencies:

   cd frontend
   npm install

3. Run Development Server:

   npm run dev

4. Access: http://localhost:5173

Important

Manual frontend requires the Backend (Docker) to be running on http://localhost:8000.

---

📸 Usage & Calibration

To get accurate results, follow these steps:

1. Upload: Select a top-down clear photo of your coffee grounds. Place a reference object (like a coin) next to them.
2. Calibrate:
   • Click and drag a line over the reference object (e.g., the diameter of a NT$10 coin).
   • Enter the real-world length (e.g., 26 mm for a 10-dollar coin).
3. Analyze: Click "Run AI Analysis".
4. Result: The AI will identify all particles and calculate their average diameter in millimeters.

---

🧹 Maintenance

The system includes an automatic cleanup mechanism:

• Uploaded images are automatically deleted from the server immediately after AI analysis is complete to prevent storage accumulation.

---

👨‍💻 Author

Ricepaste/coffeegrindsize (Modernized Version) Enhanced with Deep Learning and Microservices for Future Resume Enhancement.