Semantic segmentation

Segmentation Services for Computer Vision Training Data

Semantic segmentation annotation helps train computer vision based AI models by assigning each pixel in an image to a specific class of object. Instance segmentation annotation adds further detail to training imagery by separately labeling objects belonging to the same class.

Keymakr provides semantic and instance segmentation annotation to suit the needs of any machine learning project. Our experienced, in-house teams of annotators are able to accurately apply segmentation techniques to images and videos at an affordable price.

Semantic segmentation annotation

Annotations can have different precision levels up to "pixel perfect". All the projects are managed by machine learning experts that understand exactly how your data should be labeled. As a result, we offer high-quality and consistent data for your machine learning models.

Semantic Segmentation for Computer Vision AI

Computer vision is the process of computers gaining high-level understanding of digital images and video.

Semantic segmentation annotation is one way in which machine learning models are able to gain this understanding. To create semantic segmentation imagery data annotators use annotation tools to divide images into sets of pixels that correspond to real-world categories.

By using polygon labeling techniques annotators can trace the outline of each relevant class. For example an image of a dog herding sheep in a field would be separated, by pixel, into the classes: sheep, sheep dog, field, and sky.

This training data allows AI models to precisely locate objects in an image and distinguish between classes of objects. This process can be applied to individual images or the multiple frames of video data. Semantic segmentation has applications for a number of industries.


In order for AI models to distinguish between types of clothing items they need to be trained with semantically segmented image data.

Using annotation tools we can assign a colour to the pixels which make up each item of clothing in an image. This colour corresponds to a class of clothing item: shirts, shoes, dresses.These annotated training images allow for the creation of virtual wardrobes that allow customers to try on clothing through their computer screens.

Autonomous Vehicles

Safety is paramount when it comes to autonomous vehicle AI. Self-driving cars need to be able to navigate complex environments and avoid dangerous situations.

To help them achieve this Keymakr provides annotated video training data to developers. In these videos, semantic segmentation divides the road from the sidewalk, and vehicles from pedestrians; therefore ensuring safety for all road users.

Instance Segmentation for Computer Vision AI

Instance segmentation is a subset of image annotation that adds additional levels of labeling detail. Instead of dividing each pixel into defined classes, instance segmentation identifies each instance of each object appearing in a given image.

In the case of the sheep herding image, instance segmentation means that each individual sheep will be outlined and labeled. The AI model is therefore able to recognise where one sheep ends and the other begins. This added granularity is vital for training data across a range of industries.

Waste Management

The careful management of waste and the recycling process is vital for protecting humans and the environment. Instance segmentation of training images teaches waste management AI to classify waste into real world categories, whilst also recognising the quantity of each individual waste type.

Many waste items can be harmful to humans and should not be handled. AI models can quickly identify and isolate dangerous materials, and prevent harm to human operators.


Security cameras provide vital information for law enforcement and can also help to prevent crimes from occuring.

However, they need to be continually monitored in order to work most effectively. Computer vision based AI has the potential to provide this constant presence.

Keymakr provides instance segmentation annotation for security video training data. Instance segmentation utilises polygon labeling to zero in on important objects, for example at backpack or a gun. Using this training footage security AI is able to spot crimes in real time and alert law enforcement.

Let's Talk

Thank You.

We will contact you shortly!


We encountered an error submitting your form. If you continue to experience this issue please contact us directly at

I have read and agree to the Privacy Policy