What is polygon annotation?
Polygon annotation is a precise way to annotate objects by selecting a series of x, y coordinates along its edges. Polygon annotation can hence have pixel perfect precision whilst being highly flexible and adaptable to a variety of shapes.
Where can it be used?
Polygon annotation is very often used in object detection and recognition models due to its flexibility and pixel-perfect labeling capability.
- Medical: outlining the anatomy of internal organs on CT scans
- Geospatial: getting precise building footprints
- Automotive: detecting damages on cars and calculating the repair costs
- Industrial: making precise liquid estimations with centrifuge
- Agriculture: monitoring the growth of plants
- Retail: detecting products in a shopper’s basket
- Polygons are ideal for labelling irregular shapes.
- Pixel-perfect annotation means that no irrelevant pixels go inside of the annotated area.
- Polygons may take considerably longer than bounding boxes to annotate, depending on the complexity of the object.
- Not all annotation tools include the ability to make holes inside of polygons, or to signal that two polygons belong to the same object in the case of truncation.
- Some tools require users to click along the border of the object in order to create nodes connected by straight edges, while others allow for freehand tracing which can be more precise. The best tools allow users to easily toggle between the two modalities.
- Consider whether overlapping polygons must be distinguishable. If so, using layers would be recommended.
- Round shapes often require more nodes as it’s harder to achieve the smooth look and shape with fewer nodes. On the contrary, it’s best to put less nodes where the object is straightly shaped, as adding more will decrease the smoothness of the annotation.