infer-net: Dockerized Image Classification API

A production-style, end-to-end machine learning application that serves an image classification model through a containerized FastAPI web service.

This project was built to demonstrate best practices in software engineering applied to the ML lifecycle, including a Docker-first approach, automated CI/CD with GitHub Actions, configuration management, and automated testing.

Key Features

• FastAPI Backend: A high-performance API built with FastAPI.
• Dockerized Service: The entire application is containerized with Docker and orchestrated with Docker Compose for easy setup and deployment.
• ML Model Integration: Serves a TensorFlow/Keras CNN model trained on the Fashion MNIST dataset.
• Direct Image Upload: A robust /predict endpoint that accepts direct image file uploads.
• CI/CD Pipeline: Automated workflow using GitHub Actions for:
  • Linting with ruff.
  • Testing with pytest.
  • Build Validation of the Docker image.
• Configuration Management: Externalized configuration for maintainability (model path is set via environment variables).
• Automated Testing: Unit and integration tests to ensure API reliability.

Technology Stack

• Backend: Python, FastAPI
• ML Framework: TensorFlow / Keras
• Containerization: Docker, Docker Compose
• CI/CD: GitHub Actions
• Testing: Pytest
• Linting: Ruff

---

Getting Started

Prerequisites

• Docker and Docker Compose installed on your local machine.
• A trained model file (infer-net.keras) available. For CI, this is handled via GitHub Releases.

Running the Application

1. Clone the repository:

   git clone https://github.com/zhu-weijie/infer-net.git
   cd infer-net

2. Place the model file:

   • Create a directory artifacts.
   • Place your trained infer-net.keras model file inside it.

3. Build and run the container: Use Docker Compose to build the image and start the service. This command will start the API on http://localhost:8000.

   docker-compose up --build

---

How to Use the API

Once the application is running, you can access the interactive API documentation at: http://localhost:8000/docs

Prediction Endpoint

You can send a POST request to the /predict endpoint with an image file to get a classification.

POST /predict

Example with curl

Use curl to send an image file and receive a prediction.

curl -X 'POST' \
  'http://localhost:8000/predict' \
  -H 'accept: application/json' \
  -F 'file=@/path/to/your/image.png;type=image/png'

Example Response:

{
  "predicted_class": "Ankle boot",
  "confidence": 0.9985
}