Back to our work
Computer vision

A computer vision platform, from annotation to real-time applications

A computer vision platform that takes a project from data storage and annotation through training to live inference. Our team built two applications to demonstrate the full workflow.

Period
Jun 2020 - Jul 2022
Our scope
Platform architecture & development
Engagement
Delivered by the aiLabs founders
Our dataset preparation application showing different views of the same object.
Our annotation application: several views of the same physical object in one dataset.

Key results

  • The team delivered the platform to Stellar Solvers and demonstrated its full workflow with two applications it built.
  • After training on manually annotated images, the model labelled new images. Users reviewed the labels and corrected them when needed.
  • Retraining started from the UI, with the new model ready to use as soon as training finished.
  • Real-time recognition on a 40-60 fps video stream in the coffee machine augmented reality (AR) demonstration.

The challenge

Stellar Solvers wanted one working environment for computer vision projects, much like CAD software for architects. Data, annotations, models and model versions had to live in one system, and a trained model had to go straight into real-time applications, with isolated environments for different customers.

Our contribution

  • Designed the platform architecture and a multi-tenant data model: workspaces, roles and access control, datasets, annotation layers and tasks, models and model versions.
  • Developed a cross-platform desktop annotation application in C++ and Qt for Windows and Linux, with AI-created annotations that users approve or correct.
  • Built training and auto-annotation services, so a model trained on the first manual annotations labelled new images and could be retrained from the UI.
  • Built low-latency inference pipelines: browser video over WebRTC and WebSocket, Go connectors, NNG messaging and C workers running YOLO detectors.
  • Added text detection and optical character recognition (OCR) on packaging and a word-based product classifier, used to onboard and recognise new products.
  • Developed the backend API, authentication with JWT and signed requests, and web applications for recognising objects and onboarding products.
  • Built two applications on the platform. The spice recognition app tells near-identical jars apart by the words on their labels, using a classifier trained on hundreds of annotated images. The coffee machine AR app recognises the machine in an ordinary camera feed and tracks feature points for a steady overlay.

What the work delivered

Using the two applications, the team demonstrated the full workflow step by step, from preparing and annotating a dataset through training a model to running recognition on live video.

More engineering stories

View all work
Applied ML

Predictive learning analytics

A service that trains a model on each institution’s data, explains its predictions and analyses text, topics and sentiment. It launched in 2021 and was extended in 2025-2026.

Data engineering

Enterprise data warehouse

A BigQuery warehouse that brought sales, e-commerce, web analytics and sales-force data together, with Power BI reporting and market basket analysis for the US and Canada.

Applied ML

Real-time bidding optimisation

A bid optimisation system for advertising campaigns on The Trade Desk: event data pipelines, simulation-tested algorithms and services that adjust bids across many dimensions every hour.

Let’s build something useful

Bring us the
hard part.

A new AI capability, a complex data problem or a platform that needs to evolve. Tell us what you’re working on.

Discuss your project

Start a conversation

info@ai-labs.tech

aiLabs Innovations FZ-LLC
Compass Building, Al Shohada Road, Al Hamra Industrial Zone-FZ, Ras Al Khaimah
United Arab Emirates
Licence No. 47014246

For an initial conversation, share your goal, the current system and the challenge you want to solve.