FastAPI Image Cutline Service

I need a clean, self-contained FastAPI microservice that accepts an uploaded JPEG or PNG, automatically traces the outer contour of the visible subject, expands that shape by a configurable offset, and returns the resulting cutline as an SVG file. Auto-outline detection is the single most important processing step; edge smoothing and colour tweaks are nice-to-haves but not required for this phase.

Here’s the flow I have in mind. A user hits /outline with a multipart upload, the API queues the image, runs the outline routine (OpenCV, scikit-image, or any library you prefer), inflates the path by the requested number of pixels or mm, then streams back plain-text SVG. A second, optional endpoint /health is enough for monitoring. The service must comfortably handle images up to roughly 30 MB without timing out.

I’ll review the work against a small set of deliverables:

• Fully-commented Python code built on FastAPI, uvicorn-ready
• Requirements.txt or pyproject.toml listing every dependency
• A sample request/response JSON and the generated SVG file
• One annotated screenshot that shows your script outlining a real image and the offset path it produced
• Concise README covering setup, environment variables and a quick-start curl example

Containerisation (Docker) is welcome but optional; if you include it, keep the image lean. Please name any third-party model weights or licences you plan to use up front so I can clear them internally.

Looking forward to seeing your approach and the example result.