Systems
Training Data Storefront
Enterprise-scale, multi-source data discovery, ontology and label management for your AI/ML workflows. Browse and filter overhead imagery by provider, sensor modality and image quality, then assemble and label the datasets your models need.
How it works
One place for the whole data workflow
TDS is a full-featured web app, blending automated ingest of data, a familiar browse/search/filtering user experience, a collaborative annotation interface, and powerful data extraction tools.
Assemble reference datasets with precision
TDS’s Dataset Builder supports the incremental assembly of reference datasets, while providing visual analytics into dataset biases. This insight helps inform prioritization for subsequent data harvesting and curation.
Integrated “Segment Anything” models promote the creation of high-precision polygons while minimizing repetitive mouse-clicks.
Unify your data ecosystem
A companion app to TDS, the Ontology Translation Service (OTS), helps inform intelligent class labeling decisions when creating the ontology for your mission.
Mapping your classes to other reference ontologies promotes discoverability and interoperability within the ever-growing ML ecosystem.
Streamline curation and collaboration
Find the exact data you need with a familiar browse, search, and filtering experience.
A built-in collaborative annotation interface allows your team to work together seamlessly, enriching datasets in real-time.
Core capabilities
Purpose-built tools for common AI workflow challenges
Automated data ingest
Seamlessly pull from diverse data sources into a unified storefront.
Incremental dataset assembly
Build and version your datasets piece by piece for maximum control.
Visual bias analytics
Gain crucial insights into dataset imbalances before you train.
AI-assisted annotation
Use “Segment Anything” models to create high-precision polygon labels in seconds.
Intelligent class labeling
Let the Ontology Translation Service (OTS) guide your labeling decisions.
ML ecosystem interoperability
Promote data discoverability by mapping your classes to reference ontologies.
Inside the storefront
A familiar interface over an unfamiliar problem
Browse, mapping and annotation in one web app, so curating training data looks like using a catalogue rather than running a pipeline.
Browse, search and filter
Find the exact data you need by provider, sensor modality, image quality and collection geometry, then add it straight to a dataset.
Map classes between ontologies
OTS relates your mission classes to reference ontologies, so a label means the same thing to the next model that consumes it.
Inspect every annotation
Each label carries its own provenance — who uploaded the source image, when it was adjudicated, and which collection it came from.
Accelerate your AI workflow
Discover the benefits of TDS
For pricing information and more product details, please contact our team. We'll provide a personalized overview to show how TDS can integrate with your mission.