SamaHub Platform

Deploy Machine Learning Faster

SamaHub is Samasource’s training data annotation platform used to manage the entire annotation lifecycle. The platform ensures high-quality, human-powered training data and validation for world leading AI teams.

Data labeling for machine learning

Vector Annotation

Annotate 2D images, video and 3D objects with efficiency using cuboids, bounding boxes, polygons, lines and more.

Semantic Segmentation

Segment images with raster masks using paint, fill and superpixel segmentation tools to add pixel-level labels.

Sensor Packages

Visualize and label 3D LiDAR and radar point cloud data and synchronize ground truth data across sensors.

Dynamic Labeling

Label and track objects across multiple frames with automatic interpolation for video annotation.

About Samahub

SamaHub’s secure, cloud annotation platform delivers high-quality training data at scale.

Interested in learning more about our platform? Download our solution brief:

SAMAHUB ANNOTATION SUITE

Introducing AI Starter Kit

On-demand training data for machine learning.

FAQ

What is SamaHub?

SamaHub is a web-based data labeling platform for machine learning. SamaHub has three main functions: task distribution, data structuring (training data with annotations, data categorizing, etc.) and data quality management.

How can I get started with SamaHub?

There are two ways to get started with SamaHub:

  1. For established AI teams, request a demo of SamaHub here.
  2. New to AI or have research and development needs? Explore SamaHub’s training data on-demand service.
What types of annotation does SamaHub support?

SamaHub supports various types of training data annotation for 2D images, 3D point clouds, videos and sensor data. The annotation suite includes automated interpolation for video annotation, semantic segmentation with raster masks, support for 3D lidar and radar data, sensor fusion, cuboids, bounding boxes, polygons and lines.

What file formats can SamaHub deliver training data?

Your training data can be delivered in JSON, XLS and CSV, or delivered via API. Contact us if you don’t see your data type.

What input data formats does SamaHub support?

For text, SamaHub supports strings, integers and dates.

  1. String - allows any combination of numbers, symbols and characters. Long strings (up to 16MB) are permitted.
  2. Integer ­- allows any 32­bit integer. Negatives permitted.
  3. Date ­- allows dates in ISO­8601 format (YYYY­mm­dd). All other date formats are not permitted.

For assets, SamaHub supports images, videos and point cloud sequences.

  1. Images - .jpg, .jpeg, .png.
  2. Video - .mp4. Videos are converted into .jpg frames for processing. Folders of images, as well as images in sequence in one folder are also supported.
  3. Point Clouds - .pcd, .las, .laz, .ptx, or other formats with an (x,y,z) set of coordinates.

If you don’t see your data type, contact us.

Does SamaHub have an API?

Yes, SamaHub’s platform API and webhook is available to manage your project, tasks and data. You can create tasks and send task assets, prioritize tasks in the work queue, delete tasks and assets, reject tasks that need more work, and receive deliveries of training data.

What can I expect from SamaHub project workflows?

SamaHub’s workflow management features allow you to customize instructions and annotation labels, and edit workflows as needed, as your algorithm learns. You can also use workflows to check data quality by individual annotator, team or project and request rework of data as needed.

Can I add training data that already has pre-labeled annotations into SamaHub?

Yes. You can import training data with annotations into SamaHub using our API. Data can be edited and updated.

Where can I find information on SamaHub feature releases?

Visit our blog for information on the latest features and updates.