Starlette API Server to Classify Bears

A dockerized Starlette ASGI server that can classify an image as Black, Grizzly or Teddy bear.

Just following the excellent and fun Lesson 2 of FastAI Course1 v3 on exposing a trained image classifier model behind a public API. Also the design and code is shamelessly copied from Simon Willison’s Courgar-or-Not.

A running server is currently deployed to Render.com using FastAI student’s credit as of Sep 20th 2019, but no guarantee it will still be running by the time you read this.

How to Run and Test

• Clone the project, then run it by:
  • Using Docker
    • (from parent directory of bguan-bear…)
    • docker build -tbguan-bears bguan-bears
    • docker run -p 8008:8008 bguan-bears
  • or using a Python 3.6+ environment with FastAI and Starlette installed
    • (cd into bguan-bears…)
    • python bears.py serve
• Point browser at https://localhost:8008 (if running locally) or https://bguan-bears.onrender.com (using my deployed demo server that may be running) and upload photo or submit URL pointing to image
• You can also programmatically hit the HTTP API directly, e.g.
  • https://bguan-bears.onrender.com/classify?url=https://n.nordstrommedia.com/id/sr3/8469a1d0-a660-49df-b4e5-96deb40cdbaf.jpeg

Note:

• To protect server, image upload or submit by URL is limited to 5MB