DalšíDalší průvodce
Segmentace lékařského obrazu pomocí nnU-Net
Vizuální AI
Vizuální průvodce AI
Polygon and mask annotation is the process of outlining the exact pixels that belong to an object in an image, rather than just drawing a box around it, so a model can learn where an object's boundary actually falls.
Poslechněte si tento návodAI-vyprávěný · ~3 min
It matters because tasks like medical imaging, autonomous driving and photo editing need pixel-level precision, not just approximate location.
Segmentation annotation comes in two closely related forms: polygon annotation, where a labeler clicks a series of vertices to trace an object's outline, and mask annotation, where every pixel belonging to an object is painted or selected directly. Polygons are stored as coordinate lists and are compact and editable, which makes them common for objects with fairly smooth edges like cars, buildings or road markings. Masks are stored as pixel-level bitmaps and are better suited to objects with irregular or fine-grained boundaries, such as hair, foliage, or smoke, where a polygon with straight edges between vertices would blur the true shape. Manual tracing can be time-consuming for complex objects, and quality checks may include a second reviewer checking boundary accuracy. Promptable segmentation models, including Meta's Segment Anything Model (SAM), released in 2023, changed some workflows: in supported tools, a person can click a point or draw a box and receive a proposed mask, then review and correct it. The amount of time or cost saved depends on the task, model, tool, and proposal quality. A common misconception is that segmentation always means separating an object from its background; in practice, panoptic segmentation labels every pixel in an image, including background regions like sky, road and grass, not just discrete foreground objects. Another misconception is that masks are strictly more accurate than polygons; for polygon-friendly shapes, a well-placed polygon can match mask accuracy while using far less storage and being easier for a human to verify.
A self-driving car dataset team traces the exact outline of pedestrians, curbs and lane paint so a segmentation model can tell drivable surface from sidewalk down to the pixel.
A radiology labeling team paints masks over tumors in CT slices, marking the boundary voxel by voxel so a diagnostic model learns the tumor's true shape rather than a rough box.
A photo-editing app's background-removal feature is trained on masks where annotators traced hair strands and clothing edges, so cutouts do not leave a hard rectangular halo.
A satellite imagery company has annotators draw polygons around individual building footprints and crop field boundaries so a land-use model can count structures and estimate farm plot sizes.
Promptable segmentation models are likely to keep pushing annotation effort from tracing toward verification, with humans increasingly correcting model proposals rather than drawing from scratch. This should lower the cost of building segmentation datasets for narrower domains, such as specific medical imaging modalities or industrial inspection, where general-purpose models still make systematic errors on unfamiliar shapes or textures. Video segmentation, where masks must stay consistent across frames as objects move, remains harder to automate fully and will likely continue needing more human review than single-image segmentation for some time.
Newsletter zdarma
Tři ověřené příběhy ze světa AI každý všední den ráno, psané srozumitelnou angličtinou. Navždy zdarma, bez reklam.
Jeden e-mail každý všední den. Odhlásit se můžete jedním kliknutím. Vaši adresu nikdy neprodáváme ani nesdílíme.
Otestujte se
Okamžitá zpětná vazba ke každé odpovědi a sdílitelný certifikát s ověřitelným ID po absolvování kurzu.
Podpořte bezplatnou AI vzdělávání. AI Understanding je nezisková organizace 501(c)(3) — žádné reklamy, žádná platba za přístup, nikdy. Darujte
Polygon and mask annotation is the process of outlining the exact pixels that belong to an object in an image, rather than just drawing a box around it, so a model can learn where an object's boundary actually falls. It matters because tasks like medical imaging, autonomous driving and photo editing need pixel-level precision, not just approximate location.
Polygons connect vertices with straight lines, which poorly represents fine, irregular boundaries like individual hair strands, so pixel-level masks capture that detail better.
The original SAM is a promptable segmentation model; its proposed masks still require review and its support is version-specific.
Panoptic segmentation covers the whole image, background included, rather than isolating discrete foreground objects only.
Polygons store an ordered sequence of coordinate points that define the boundary path, which is later rasterized into a mask when needed.
RLE is a compact way to store masks by encoding consecutive runs of the same pixel value, which is efficient since large mask regions are often uniform.
Učte se dál
Pro toto téma bylo vybráno více průvodců
DalšíDalší průvodce
Segmentace lékařského obrazu pomocí nnU-Net
Vizuální AI