2D Bounding Boxes
Tight rectangles around every subject in a frame — the foundational label behind object detection, retail vision and safety AI.
VisionDetection
Services
AI annotation, training datasets and the operations behind them.
A nine-service annotation console for modern AI teams, plus the surrounding data operations — collection, cleansing, coverage analytics and programme support — that keep every labelled dataset honest end to end.
The flow
Raw records in. Structured, verified data out.
Every engagement runs through the same disciplined flow — processing, gears of technology-enabled work, human review where it matters, then structured records the client can actually use.
01 · Annotation types we ship
Ten label types. One annotation console behind every dataset.
From 2D boxes to LiDAR point clouds, multilingual NLP and document OCR — the label formats we deliver as production-ready training and evaluation data for modern AI teams.
Polygon & Semantic Segmentation
Pixel-perfect outlines and per-pixel class masks for medical imaging, autonomous driving and precision retail applications.
VisionPixel-level
Keypoint & Pose Estimation
Skeleton-graph labels for sports analytics, ergonomics, fitness apps and human–robot interaction — every joint tagged and linked.
VisionKeypoints
3D Cuboid Annotation
Volumetric boxes with width, height and depth for AR overlays, warehouse robotics and driver-assistance systems.
3DSpatial
LiDAR Point-Cloud Labelling
Per-point class, instance and cuboid annotation on raw LiDAR scans — the ground truth self-driving and robotics stacks are trained on.
LiDARAutonomy
Video Tracking & MOT
Frame-by-frame object tracking with stable IDs across occlusion — behaviour analytics, sports vision and multi-object tracking.
VideoTracking
NLP & Named-Entity Recognition
Entity spans, relations, intent and coreference — the linguistic labels behind search, chat, extraction and compliance NLP.
TextNLP
Audio Transcription & Events
Word-level transcripts, speaker turns, timestamps and sound-event tags for voice assistants, contact-centre AI and ambient audio.
AudioSpeech
Sentiment & Classification
Topic, intent, sentiment and moderation labels applied at scale with disciplined guideline enforcement and inter-annotator checks.
TextClassification
OCR & Document Field Extraction
Text detection, key-value field capture and clause labelling on invoices, forms, contracts and records — accuracy under real-world document variation.
DocumentsOCR
02 — 05 · Broader capability
Around every labelled dataset — the operations that make it work.
Sample preparation, annotation analytics, versioned training sets and programme support — the surrounding work that turns Section 01's labelling into production-ready training data.
Group 02
Data Operations
Everything that has to happen before a sample is ready to annotate — sourcing, entry, cleansing, conversion and validation of the raw material that feeds the labelling console.
Data collection
Data entry & transcription
Data cleansing & normalisation
Format conversion
Data validation
PII redaction
Deduplication
Sample preparation
Group 03
Annotation Analytics & Reporting
The numbers behind the labels — coverage, class distribution, inter-annotator agreement, throughput and quality trends — reported clearly enough that decisions get made.
Annotation coverage analysis
Class distribution reports
Inter-annotator agreement
Throughput dashboards
Quality trend reporting
Delivery reporting
Group 04
Training & Evaluation Datasets
Assembled, versioned datasets that ship straight into your training and evaluation pipelines — training splits, gold sets, HITL corrections, edge-case packs. Section 01 is the bench that produces them.
Training splits
Evaluation gold sets
Preference & ranking data
HITL correction sets
Edge-case packs
Dataset versioning
Group 05
Programme Support
The programme layer around an annotation engagement — guidelines, taxonomy management, reviewer onboarding, documentation and delivery co-ordination. The quiet work that keeps a labelling operation honest.
Annotation guideline authoring
Taxonomy & schema management
Reviewer onboarding
Documentation & runbooks
Delivery co-ordination
Programme reporting
Scope
We only publish services we can genuinely deliver.
If your requirement sits at the edge of what's listed — or spans two groups — start a conversation. The team will tell you plainly whether we're a good fit and, if not, what a good fit would look like.
Discuss your project