Is a labeling method that marks a linear feature in an image or video using a series of connected straight line segments, rather than a box or a filled in shape. It's used whenever the thing being labeled is fundamentally a line, a road edge, a lane marking, a pipeline, rather than an object with a clear boundary all the way around it.

‍Why it matters

A bounding box or segmentation mask works well for objects that take up an area, but a road lane marking or a power line doesn't really have an inside, it's just a path. Polylines let annotators trace that path directly, point by point, which gives a model the actual shape and direction of the line instead of forcing it into a boxy approximation that would include a lot of irrelevant background.

‍How it's labeled

Annotators click a series of points along the feature to build the line, usually working zoomed in to place each point accurately rather than approximating the path with just a few clicks. Tools like CVAT and Labelbox support polyline annotation directly alongside boxes and segmentation masks. Because the accuracy of a polyline depends on how carefully each point was placed, this type of annotation tends to need more review than a simple bounding box would.

‍Where it's used

Autonomous vehicle systems rely on polyline annotation constantly for lane detection, tracing exactly where lane markings run so the car knows where it's allowed to drive. Mapping and GIS work uses the same approach for roads, rivers, and property boundaries. Infrastructure inspection is another common use, tracing power lines or pipelines in aerial or drone imagery to monitor their condition over a large area.