Polygon annotation
Polygon annotation is an annotation type that outlines an object by placing a series of connected points around its boundary, producing a closed shape that follows the object far more tightly than a bounding box. It is a common way to create segmentation masks by hand.
In this article
Building visual or physical AI?
Let's talk.
What is polygon annotation?
Polygon annotation traces an object's outline as an ordered set of vertices connected by straight edges. Because the shape conforms to the object's contour, it captures much more of the true boundary than a rectangle, and it rasterizes directly into a segmentation mask. Annotators add, move, and delete vertices to fit the edge, more points for curvier objects. Each polygon carries a class, and multiple polygons, or polygons with holes, can describe complex or disconnected shapes.
Key takeaways
- A polygon outlines an object with connected vertices, capturing shape a box cannot.
- It is the manual route to a segmentation mask and converts directly into one.
- More vertices means a tighter fit but slower labeling, a cost-versus-precision trade-off.
Types and variations
- Single vs multi-part: one polygon, or several for disconnected regions.
- With holes: an outer ring minus inner rings for objects with gaps.
- Manual vs model-assisted: hand-placed vertices, or generated by click-to-segment and then edited.
How it works
Annotators click around the boundary to drop vertices, then nudge them to fit, increasingly starting from a model-proposed outline.
Polygon vs box vs mask
| Term | How it represents the object | Cost |
|---|---|---|
| Bounding box | A rectangle | Low |
| Polygon | A vertex outline of the boundary | Medium |
| Segmentation mask | Per-pixel membership | High |
Why it matters
Polygons are the practical middle ground: much of a mask's precision at a fraction of the effort. Vertex count is a quiet quality lever. Too few points and the polygon cuts corners off curved objects, a systematic labeling error baked straight into the ground truth, too many and labeling cost balloons for little accuracy gain. It is also why click-to-segment changed polygon work, it produces a dense, well-fit outline in one click that a human only has to correct.
Frequently asked questions
What is the difference between a polygon and a bounding box?
A box is a rectangle marking location, a polygon traces the object's actual outline.
Is a polygon the same as a segmentation mask?
A polygon is the vector outline, it rasterizes into a mask, which is the pixel version.
When should you use polygons instead of masks?
When you want shape accuracy with less effort than painting every pixel, or a vector format that is easy to edit.
Related terms
Learn more
See how one click now produces the outline you used to place vertex by vertex in Click-to-Segment: Interactive Segmentation in FiftyOne.
Director of Growth
Jesse Mostipak is the Director of Growth at Voxel51, where the work is helping humans find and trust what the brand knows, and teaching the Google knowledge graph and the LLMs answering on their behalf to do the same. That question, how knowledge gets built inside a system, is one Jesse has been chasing for years. Earlier versions of it ran through a New York City high school science classroom, data science and machine learning, and developer relations at Kaggle, Posit (formerly RStudio), and Baseten. The answer doesn't change much depending on whether the learner is a teenager, a software engineer, or a knowledge graph. Jesse holds a Master's in Education from CUNY Hunter College. LinkedIn
See all articles by Jesse Mostipak
Last updated November 3, 2026
Building visual or physical AI?
Let's talk.
© 2026 Voxel51 All Rights Reserved