Image Storage System

This repo consists of an image storage system implemented as a containerized Flask app with Celery for batch processing and Mongo database. Images are stored in the server's disk space, and the database stores the image's path and image metadata such as timestamp, store name, camera ID, and barcode. Upon upload, async saving of the images and their metadata is performed as a Celery task and endpoint response is immediate.

Basic endpoints, documented below, are implemented to:

• Upload a batch of images
• Retrieve batches of images (image metadata and URLs) in a paginated fashion
• Download an image given its path.

Please refer to the following diagram:

Key Files

• docker-compose.yaml defines services to launch (Flask, MongoDB, a Celery worker, and Redis)
• flask/app.py defines Flask app, database models via MongoEngine, endpoints, basic config, etc.
• test.py uploads 3 images to server

Endpoints

• POST /upload_image_batch

  JSON body:

  [ 
    {
      "file"      : "<base 64 encoded image>"
      "timestamp" : "<timestamp>",
      "store_name': "<store name>",
      "camera_id' : "<camera ID>",
      "barcode'   : "<barcode>"
  
    },
    ...
  ]
  

  See tests.py for example.

• GET /image_batches?page=<page>&per_page=<per_page>

• GET /uploads/<uuid>.jpg

Database

The database has a simple batch structure:

[
  [
    [ 
      {
        "file_path" : "<relative image URL>"
        "timestamp" : "<timestamp>",
        "store_name': "<store name>",
        "camera_id' : "<camera ID>",
        "barcode'   : "<barcode>"
      },
      ...
    ],
  ...
  ]
]

Testing

• docker-compose up
• python3 test.py
• curl http://localhost:5000/image_batches\?page\=1\&per_page\=100
• curl http://localhost:5000/uploads/<uuid>.jpg

Future

A practical production implementation would probably store images in separate cloud storage such as Amazon S3. Other backend framework alternatives to Flask might include FastAPI or Django REST.

To deploy this as is, at least a production web server should be stood up bridging to Flask with uWSGI or Gunicorn middleware in addition to version freezing, better error handling, and logging. Additional endpoints could be added for retrieval/querying by metadata.