Pixel Perfect Image and Video annotation Services
We are a leading annotation service provider thanks to our proprietary annotation platform combined with a professional in-house annotation team.
We specialize in video annotations and create consistent high-quality data for your machine learning models.
Our platform supports complex tasks such as object tracking on multiple videos and attribute hierarchy. We process videos of any size by using bounding boxes, points, lines, polygons, and multi-segment lines to markup video frames.
We create superior quality data that is backed by excellent customer service. We work with you to find the best strategy to approach your project. By combining advanced tools with in-house professional annotators, we guarantee the incredible results.
You no longer need to crawl the web, manage a crowd of workers, or contact individual data vendors yourself.
We collect the data for you. If the data doesn't exist, we will create it in our production studio.
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.
Keymakr's technology allows to easily annotate videos of any length and size with 99.9% precision. The annotation may consist of, but not limited to:
- Object recognition and tracking in multiple frames.
- Tracking of numerous objects
- Object recognition by attributes: ID number, license plate, mugshot of a person, or any other attribute.
- Tracking of multiple attributes per object while enabling the possibility of changing the attributes on the "key-frame".
- Tracking of an object through all frames of the video by a certain attribute such as a license plate or a face mugshot.
Objects may disappear from the frame and re-enter it later on, this fact will be detected, and the tracking will continue with the same Object ID and attributes previously assigned.
Each object on the video frames can have multiple attributes assigned to it.
- Changing the value of the attribute on a "key frame". "Key frame" means the frame were any attribute or parameter changes. It can happen at any time and only the relevant attribute or parameter is adjusted on the "key frame" while the rest remain the same.
- While the object's ID and the rest of the attributes stay the same, we continue to track the object with the new attribute until the next "key frame", when another attribute changes.
Challenge: Identification of specific people by security camera
20 million images of people from various ethnicities were collected and annotated by our team. Face landmarks and attributes were marked for each image: eyes, nose, brows, emotions, etc. As a result the model was able to recognize individual faces from different angles, even in a crowd.
Challenge: learning how to track in-car behavior of the driver and passenger.
500 hours of in-car video footage of various people driving were collected and annotated. 20,000 images of streets from the US and Europe's major cities were segmented with a pixel-perfect precision: cars, trees.
We have developed proprietary tools and techniques for a faster and more efficient data collection process. Doing so has reduced the time and increased our capacity for mass collection of data.Download pdf
Challenge: Brain tumor recognition by AI.
For this medical project, we provided the client with overall 3500 MRI scans - head sections with brain tumor and those with a healthy brain. All the images were annotated by certified radiologists who specialize in tumor diagnosis.
Keymakr specializes in tailor-made solutions. If your data cannot be found anywhere, we can set up a production studio and create the data uniquely required for your project. For example, we can create footage of people driving a vehicle, videos of busy highways, video footage of workers in busy warehouses, etc.