Semantic image segmentation is the division of an image into separate groups of pixels. Each group refers to a specific object and is outlined, creating a color mask.

Semantic image segmentation

Go back

What goals can you achieve by using image segmentation

SERVICES → LABELME.RU

Image segmentation is an integral part of computer vision annotation, affecting an infinite number of applications. We have extensive work experience and professional staff, who regularly improve their knowledge and learn new image annotation technologies. We will be happy to take your order, discuss the details and prepare a high-quality dataset in a short time.

Technical assignment template

Types of segmentation

Image matting

Suitable for those who need to annotate a specific object in different images. When small details merge with the background, our specialist can use values from 0 to 1 for similar areas using a tool such as Adobe Photoshop.

Multi-class segmentation

This type of segmentation is suitable for you if you need to annotate many objects in the image. Under this condition, we observe the strictness of the object contour in the image and its belonging to a specific class. Our specialist labels it using closed polygons during segmentation.

Scene parsing

This method works for you if you need to assign a particular class to each pixel in the image. In this case, we perform partitioning with unclosed polygons and Bezier curves, refining the contours of the objects, and apply area filling. In the output we get datasets with images completely covered by class masks with non-overlapping boundaries.

Application examples

The LabelMe team is working on data labeling for a wide variety of businesses:

Mobile app development

We perform data labeling for various tasks of mobile applications. These tasks include detailed segmentation of photos and videos from front-facing cameras for image detection and processing. We annotate the following imperfections of the skin: wrinkles, eyebags, sensory organs, face shapes, and much more.

Medical field

We attract the best doctors to annotate all types of medical data. Our specialists pay great attention to the borders between fabrics and select the most optimal tools. For example, polygonal cutting - to remove only the necessary areas. Using pixel-by-pixel segmentation, we can annotate even the smallest objects with high accuracy.

Product defect control

We'll carefully study the types of defects and select the most accurate and effective way of labeling. Thanks to this, we detect all dents, chips, kinks, fractures, seam deformations, smudging of paint, etc. During the annotation, performers focus on the edges of defects and can segment damages of arbitrary shapes.

Agricultural sector

We'll segment the images of plant crops at different stages of maturity, considering the sizes, shapes, and colors based on technical specifications. We can also segment field images highlighting weeds to identify weeds and predict yield or damage. We do the pixel-by-pixel labeling of problem areas on images from drones or satellites.