ServicesManaged data operationsData annotation & validationDedicated teamsAll services
TechnologyIntelligent workflowsModel orchestrationQuality assurance
Applied AIPoint AI Studio
IndustriesDefense & SecurityAutomotive & MobilityRobotics & Industrial AIAll industries
CompanyAbout usMissionOur team
Discuss your project

Home / Point AI Studio

Point AI Studio

Your data.
Connected, organized and accessible.

A workspace for managing, reviewing and using your AI data. Our team does the annotation inside it; your team organises, checks, trains on and exports the results.

Explore Studio with us See how we work
aerial_survey · v12 · 48,200 items
Class
car31,402
truck4,118
vessel1,206
person8,950
Sensor
EO40,100
IR8,100
Review status
Approved36,800
Raised for review212
Altitude (m)
120480
car · 27
car · 2
lidar
mri
satellite
vessel · 4
container
solar
1,240 of 48,200 selectedCreate subsetExport subset

Our team does the annotation. Your team uses the data.

Point AI Studio is not a self-service labelling tool. It is where our managed data operations happen, and where datasets, reviews, quality and exports stay together instead of travelling by email or shared drive.

Point AI’s team

Labels, corrects and reviews

Our trained, employed annotators correct model pre-labels, raise unclear cases as questions and review each batch before delivery.

Your team

Manages, reviews and uses the data

You organise and filter datasets, review labels, follow quality batch by batch, train models and export.

Data management & curation

Organise and find.

Projects, datasets and label versions in one workspace, with the filters to pull out the subsets that matter.

  • Image, video, 3D and DICOM, with every label version kept
  • Filter by class, metadata and quality
  • Build subsets for labelling, review or export
datasets
Illustrative list of datasets with modality, size and version
DatasetModalityItemsVersion
aerial_surveyimage · EO48,200v12
harbour_fmvvideo · FMV312 clipsv4
city_lidar3D · LiDAR9,800 sweepsv7
thorax_ctDICOM · CT1,140 studiesv3
Review & model-assisted labeling

Review in the browser.

A model pre-labels each item, our annotators correct it, and you see the result with its annotations.

  • Viewer for image, video, 3D and DICOM
  • Comment, approve or reject
  • Questions our team raised, with the decision
review · frame 218
Comments
Raised for reviewObject 28: person or trolley? Too small to call at this altitude.
DecisionLabel as person when a head is visible; otherwise “unknown”.
ApproveReject
ImageVideo3DDICOM
Quality dashboards

Track quality, batch by batch.

Follow each batch through review and see what was found and corrected.

  • Batch status and review progress
  • Error types found in review
  • Agreement and raised questions
quality · aerial_survey
Items reviewed
18,500
Raised questions
3
Batches delivered
4
Batch status
b_04b_05b_06b_07b_08b_09
deliveredin reviewraised
Error types found in review
missed object42%loose box31%wrong class18%occlusion rule9%
Model training & exports

Train and export.

Train models on the dataset versions we deliver, or export any batch or subset to your own pipeline.

  • Train on labelled versions
  • Download, API or cloud delivery
exports · batch_07
Format
COCOYOLOPascal VOCDOTAMOTGeoJSONKITTIDICOM SEGcustom schema
DownloadVersioned archive
Cloud deliveryDelivered to your storage
APIPull exports into your pipeline
Export batch_07

Have a challenging data problem?

Tell us what you’re building and what data you have. We’ll suggest a practical way to process, check and deliver it.

Discuss your project