Video Annotation

Video annotation for AI projects

Video annotation helps AI models to operate in fast moving, real-world environments. However, this essential process is also expensive and time-consuming. To annotate video human workers locate and label objects in every video frame. Managing this difficult task can be a significant distraction for AI company engineers and senior management. As a result many computer vision innovators choose to outsource their video annotation to professional services. Annotation providers ensure that AI projects receive exceptional video training data without the burden of management, training and quality control.

Bounding box annotation arrow

This is the most common, fast, reliable and cost-effective annotation method.

Polygon annotation arrow

This is what you need if you are dealing with irregular shapes and your project requires more precision than regular bounding box annotation.

Semantic segmentation arrow

Do you need to group multiple objects of a single category as one entity? Then semantic segmen­ta­tion might meet that need. Contact us to learn more, our specialists can help you to identify the right video annotation type for your project.

Skeletal annotation arrow

Reveals body position and alignment. This technique is commonly used in sports analytics, fitness and security applications.

Key points annotation arrow

Identify and mark key points of an object in videos, such as eyes, noses, lips, or even individual cells.

Lane annotation arrow

This technique is used for annotating roads, pipelines and rails, this is one of the annotation types most commonly used by car manufacturers today.

Instance segmentation arrow

We can easily detect instances of each category and identify individual objects within these categories. Categories like “vehicles” are split into “cars,” “motorcycles,” “buses,” and so on.

Custom annotation arrow

If your project requires a specific combination of annotation types, or even a new annotation type, we can easily achieve that for you. Alternatively, our exceptional R&D team can evaluate your project and create a completely new video annotation type based on your specific requirements.

Video annotation types

Video significantly increases the annotation workload by multiplying the amount of images that must be labeled in a dataset. Each frame of video must be as precisely annotated as an individual image. This process can be sped up by making use of object interpolation techniques. Object interpolation algorithms track a labeled object through multiple frames, allowing annotators to create video annotations much more efficiently.

The needs of AI developers determine how each frame of video training data should be labeled. Therefore, we use a variety of annotation types to achieve their desired results:

icon-1 icon-2

BOUNDING BOX ANNOTATION

icon-1 icon-2

ROTATED BOUNDING BOXES

icon-1 icon-2

CUBOID ANNOTATION

icon-b icon-bh

POLYGON ANNOTATION

icon-3 icon-4

SEMANTIC SEGMENTATION

icon-5 icon-6

SKELETAL ANNOTATION

icon-7 icon-8

KEY POINTS ANNOTATION

icon-9 icon-10

LANE ANNOTATION

icon-11 icon-12

INSTANCE SEGMENTATION

icon-13 icon-14

Bitmask ANNOTATION

icon-15 icon-16

CUSTOM ANNOTATION

BOUNDING BOX ANNOTATION

BOUNDING BOX ANNOTATION

This is the most common annotation type. Annotators use an annotation platform to drag boxes around objects in video frames.

Consequently, this technique is fast and easy to perform. However, it does not fully capture the shape of complex objects at the pixel level. AI models for autonomous vehicles often rely on video training data highlighted with bounding boxes. Labels created with this method help train computer vision models to identify objects and navigate in chaotic environments.

ROTATED BOUNDING BOXES ANNOTATION

ROTATED BOUNDING BOXES

Sometimes traditional bounding boxes do not suit particular shapes. For example an axis-aligned bounding box may work well for an image of a person standing but might not fit well with somebody laying down.

Rotated bounding box annotation allows annotators to pivot boxes so that they accurately conform to a targeted part of a video.

CUBOID ANNOTATION

CUBOID ANNOTATION

Cuboid annotation adds more dimensions to training video datasets. Annotators create cube shapes by using a 2D box to locate one side of an object, then an additional box is used to identify the opposite side.

By linking both boxes together we can create a 3D cuboid in a 2D image. This allows more information about objects to be contained in training videos, including height, width, depth, and position.

POLYGON ANNOTATION

POLYGON ANNOTATION

If you want to precisely define the shape of an object it is necessary to use this annotation type. Annotators connect small lines around the pixel outline of target objects.

Polygon annotation allows each frame of video to be segmented more accurately. This annotation type is important for agricultural management, plant monitoring and livestock management AI systems.

SEMANTIC SEGMENTATION

SEMANTIC SEGMENTATION

Environmental perception is crucial for many AI use cases. For autonomous vehicles environmental perception means having an awareness of the road environment, including road markings, obstacle locations and vehicle velocities.

By assigning each pixel in an image to a particular class semantic segmentation for video data promotes this high level situational awareness. We can label target objects and their surrounding context using polygon annotation.

SKELETAL ANNOTATION

SKELETAL ANNOTATION

This annotation type is done by adding lines to human limbs in video frames. Annotators connect these lines at points of articulation, e.g. knees, shoulders.

There are a number of AI use cases that require machine learning models to interpret the movement of the human body. As a result, video data featuring skeletal annotation often trains sports analytics systems and home fitness products.

KEY POINTS ANNOTATION

KEY POINTS ANNOTATION

Video training data can be used to make facial recognition models for security or retail applications. Key point annotation is a vital part of creating these datasets.

This technique involves annotators marking key facial features, for example mouth, nose, eyes, as they appear in each frame of video.

LANE ANNOTATION

LANE ANNOTATION

This annotation type makes it possible to label linear and parallel structures in video frames.

Examples include: power lines, train lines or pipelines. Automated vehicles rely on line annotation because it allows models to recognise road markings and stay within them.

INSTANCE SEGMENTATION

INSTANCE SEGMENTATION

This method adds granularity to semantic segmentation by indicating how many times an object appears in a video frame.

BITMASK ANNOTATION

Bitmask ANNOTATION

Keymakr’s advanced video annotation tools and our professional in-house annotation team ensure the best results for your computer vision training data needs.

Annotating videos while tracking objects through multiple frames. Each object on the video will be recognized and tracked even through different cameras or separate video segments.

CUSTOM ANNOTATION

CUSTOM ANNOTATION

Keymakr’s advanced video annotation tools and our professional in-house annotation team ensure the best results for your computer vision training data needs.

Annotating videos while tracking objects through multiple frames. Each object on the video will be recognized and tracked even through different cameras or separate video segments.

Video annotation use cases

Keymakr’s effective and affordable video annotation services support a wide variety of AI applications:

In-cabin driver monitoring

Keymakr annotated over 500 hours of in-car video footage, featuring a variety of drivers and in-car scenarios.

Annotators labeled and tracked body and facial feature movements in each video. This data allows AI models to interpret human behavior and give warnings if, for example, a driver is falling asleep.

Security AI

Skeletal annotation of video data allows AI to interpret movement. It is possible to create video data that showcases a variety of human behaviours.

Annotators use lines to identify limb positions in each frame of CCTV footage. This data is then used to train security AI models, allowing them to identify when an individual is moving erratically or behaving in a threatening manner.

play icon
SEMANTIC SEGMENTATION ON VIDEO

SEMANTIC SEGMENTATION ON VIDEO

Video data can contain additional information for AI training by deploying semantic segmentation.

This technique separates each pixel in each frame into classes of objects.

INSTANCE SEGMENTATION ON VIDEO

Instance segmentation promotes granularity in video data by identifying each individual case of a particular class of object in every frame.

For agricultural AI this might mean outlining and labeling each individual animal across an entire piece of footage.

Video annotation services, like Keymakr, also support pioneers in the field of disaster management. For example, automated drones can search large areas to find missing people, identify flooding and survey damaged buildings. These applications are trained with polygon annotation and video data.

INSTANCE SEGMENTATION ON VIDEO
VIDEO ANNOTATION WITH POINTS

VIDEO ANNOTATION WITH POINTS

Often AI models are required to identify key points in video footage. To achieve this annotators work through thousands of frames locating important positions in each image.

Video ANNOTATION WITH LINES

Line annotations are often applied to images of linear objects, such as: roads, pipelines, and electrical wires.

Keymakr produces line annotation for automated vehicle training. By locating markings and boundaries in video footage annotators help AI models to operate within the safe limits of the road system.

VIDEO ANNOTATION WITH LINES
SKELETAL VIDEO ANNOTATION

SKELETAL VIDEO ANNOTATION

Charting and analysing the movements of the human body in video is an important application for machine learning models.

Skeletal annotation reveals how the body navigates from frame to frame.

INDUSTRIES THAT WE SERVE:

icon-17 icon-18

Automotive

We can annotate videos that take place in a variety of weather conditions, in the day or the night, without missing a single detail.

icon-19 icon-20

Security

CCTV footage annotation, traffic monitoring, person or object tracking through multiple cameras and video frames.

icon-33 icon-34

Medical

Just tell us what your project requires, and we will deliver. We offer medical video collection and annotation by certified medical professionals.

icon-35 icon-36

Robotics

From delivery robots to an “eye” in the production line, we can help you do it all. Our video annotation services are here to facilitate the countless applications of robotics AI.

icon-25 icon-26

Aerial

There are many uses for drone AI, however, these models require high quality datasets. We annotate drone footage for agricultural applications, object tracking, advanced monitoring systems, disaster monitoring and management and more.

icon-27 icon-28

Agriculture and Livestock Management

We have a wealth of experience with farm drone footage annotation, real time crop monitoring and ripeness detection, as well as video footage annotation from farm cameras for livestock management and counting.

icon-37 icon-38

Waste Management

Video annotation for sorting facilities will help to identify what can and can’t be recycled without risking human exposure to potentially toxic or hazardous waste.

icon-21 icon-22

Retail

The use cases for retail AI, trained through video annotation, are numerous. For example mapping customer’s paths through the store or tracking objects from security camera footage for loss prevention.

icon-25 icon-26

Sport

We annotate videos for sport analytics and fitness applications, tracking every change in body position and alignment.


Video annotation services and tools

Video annotation is a labour intensive process that requires a significant investment of time and resources. Consequently, many AI companies looking for data annotation make use of automated annotation tools to accelerate this process. Alternatively companies can collaborate with annotation service providers, like Keymakr, and access high-quality, scalable video annotation.

Video annotation services

Annotation services, like Keymakr, provide video labeling support to computer vision AI pioneers. Keymakr's large team of skilled annotators work with proprietary technology to label video training data. In addition, this team is led by experienced managers and supported by strong quality control procedures.

Automated annotation tools

AI assisted tools can make annotation faster and easier for workers, and it is even possible to annotate entire datasets automatically. Auto-annotation tools can locate and label the same object over thousands of frames. This can significantly accelerate the video annotation process.

Video annotation services vs automated video annotation

Automated video annotation tools can create large video datasets quickly and affordably. Automation can also reduce labour costs and management pressures. Despite these strengths automated annotation can leave AI companies with a lack of support and lower quality video data.

Annotation service providers like Keymakr have a wealth of annotation experience, and Keymakr also gives AI companies access to proprietary annotation software. Annotation technology keeps data annotation tasks on schedule. In addition, outsourcing video annotation allows costs can be scaled up and down as data needs change.

Automated annotations can be cost effective but also tend to contain significant errors that impede development. Keymakr’s annotation teams work together, on-site, and are led by experienced team leaders and managers. This allows for far greater communication and troubleshooting capabilities, and ensures that video annotation quality remains at a high level.

  • Smart task distribution system. As part of the service provided by Keymakr, companies in need of video annotation have access to the unique features of Keymakr’s annotation platform. This also includes an innovative smart task distribution system that assigns tasks to annotators based on performance metrics and suitability. This helps to keep annotations precise across video datasets.
  • 24\7 monitoring and alerts. The Keymakr platform allows managers to see real time information about the status of their project. The platform can send alerts to managers when there are recurring issues with data quality or if a particular task is behind schedule.
  • Vector or bitmask. Keymakr offers both bitmasks and vector graphics for video annotation projects. Keymakr can convert both image types if necessary.
g2-bg
g2-logo

G2 Reviews

g2-logo-1 g2-logo-2
g2-star-icon
g2-star-icon
g2-star-icon
g2-star-icon
g2-star-icon

Aug 31, 2022

"Delivering Quality and Excellence"

The upside of working with Keymakr is their strategy to annotations. You are given a sample of work to correct before they begin on the big batches. This saves all parties time and...

g2-star-icon
g2-star-icon
g2-star-icon
g2-star-icon
g2-star-icon

Aug 16, 2022

"Great service, fair price"

Ability to accommodate different and not consistent workflows.
Ability to scale up as well as scale down.
All the data was in the custom format that...

g2-star-icon
g2-star-icon
g2-star-icon
g2-star-icon
g2-star-icon

Aug 12, 2022

"Awesome Labeling for ML"

I have worked with Keymakr for about 2 years on several segmentation tasks. They always provide excellent edge alignment, consistency, and speed...

g2-star-icon
g2-star-icon
g2-star-icon
g2-star-icon
g2-star-icon

Aug 05, 2022

"Great annotation service by Keymakr"

Great experience to work with Keymakr. The team is very responsive, always provides excellent results, and with perfect quality control.

g2-star-icon
g2-star-icon
g2-star-icon
g2-star-icon
g2-star-icon

Aug 22, 2022

"Great collaboration"

Keymakr is attentive to our needs making adjustments to fit our reality.
We worked together to establish an efficient information exchange and...

left-arrow
right-arrow

LET'S TALK!

Thank You.

We will contact you shortly!

Error.

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

◘ I have read and agree to the Privacy Policy


I have read and agree to the Privacy Policy
Member of Association for Advancing Automation

Member of
Association for Advancing Automation