About AnnotateIt AI
Built from real computer vision engineering experience
AnnotateIt AI was created by Yuriy Volokitin, a software and AI engineering leader with more than 15 years of experience building technical products, including computer vision and dataset tooling.
That experience is the specification. AnnotateIt is the annotation tool he wanted while doing the work: one that opens in seconds, keeps the dataset on the machine that collected it, and hands it back in a format a training script already reads.
Why AnnotateIt AI exists
Most annotation platforms are designed around a cloud service: an account, a server to deploy or a subscription to keep open, and enough annotators to justify the coordination features. For an organisation running a labelling operation, that shape is right. For one engineer inspecting a few thousand images, correcting what a model got wrong and exporting the result, it is a lot of infrastructure standing between a person and their own files.
AnnotateIt grew from the opposite end — the smallest thing that still does real work:
- Setup measured in seconds, not in deployment steps.
- Processing that happens on the machine already holding the data.
- AI assistance that removes the repetitive part without taking the decisions.
- Direct access to the media and annotation files at any point.
- Export formats a training pipeline already understands.
- A workflow sized for an individual engineer, a researcher or a small technical team.
None of that is an argument against the platforms built for larger operations — they solve a problem AnnotateIt does not attempt. It is a decision about who the product is for.
Experience behind the product
Yuriy has spent more than 15 years in software engineering — backend systems and APIs, AI integration, and delivering applications across web, desktop and mobile — including leading engineering teams building software and AI products. Part of that work was engineering leadership connected to Intel Geti, a computer vision platform, which is where annotation and dataset management stopped being an abstraction and became a daily constraint.
The specifics matter more than the job titles, and four of them shape most of what AnnotateIt does:
- Annotation quality decides model quality long before architecture does.
- Most of the real time goes into dataset management — versions, splits, duplicates, class balance — not into drawing shapes.
- Model-assisted labelling only pays off if a human can correct it faster than they could redo it.
- On-device inference is practical for far more of this work than it is usually given credit for.
Intel and Intel Geti are referenced only as part of Yuriy’s previous professional experience. AnnotateIt AI is an independent product and is not affiliated with, sponsored by or endorsed by Intel.
Product principles
Five rules the product is held to — and the reason several otherwise obvious features were deliberately not built.
Your data remains yours
Media and annotations stay in local storage on your device, and inference runs on your hardware. There is no account to create and no upload step: local annotation keeps everything in device storage, so there is nothing to send.
How local processing worksAI should assist, not obscure
Every box, mask, label and keypoint a model proposes arrives in the same editor as the ones you drew — reviewable, editable and easy to throw away. Nothing is saved as final on your behalf.
No infrastructure should be required for basic work
Starting a local project should not mean deploying Docker services or maintaining an annotation server. Install a native build, or open a browser tab.
Platforms and installsInteroperability matters
A dataset readable only by the tool that produced it is a liability. YOLO, COCO, Pascal VOC and Datumaro go in and come back out, and the round trip is documented rather than assumed.
Dataset formatsLimitations should be explicit
What the product supports, what differs between the web and native builds, and when another tool is the better answer are all written down — before you have spent a week finding out.
Platform differences
What AnnotateIt AI is — and is not
The second list is the more useful one, and it costs far less to read here than to discover in the third week of a project.
AnnotateIt AI is
- A local-first computer vision annotation and dataset tool.Privacy & local processing
- Built for detection, segmentation, classification, keypoints and video-frame workflows.Annotation tasks
- AI-assisted, and human-controlled at every step.
- Available on the web and as native Windows, macOS, iPhone and iPad builds.Download
- Suited to individual engineers, researchers and privacy-sensitive workflows.Security overview
AnnotateIt AI is not
- A managed annotation workforce — there are no annotators for hire here.
- A hosted training and deployment service. The experimental ML workflow uses a separately configured runner.
- A replacement for enterprise-scale collaborative review systems, with shared queues, assignment and per-user access control.How it compares
- A promise that every feature behaves identically on every platform — the mobile builds in particular carry a reduced feature envelope.What differs by platform
Pricing follows the same logic: currently free on every platform, with no paid upgrade or subscription.See pricing
How development is documented
AnnotateIt is not an open-source project: the public GitHub organisation carries the issue tracker and the roadmap, not the application source. Calling this “building in public” would overstate it. What is accurate is that development is documented transparently — every record below is public.
Release notes
What actually changed in each release, version by version.
Roadmap
What is planned next — and what has been ruled out on purpose.
Security overview
The security architecture and how to report a vulnerability.
Data flow
Every request the application can make, itemised, with what it carries.
Report a bug or request a feature
Public issue reporting, with templates for bugs, features and import/export problems.
Support & contact
One address, answered by the person who wrote the code.
See how AnnotateIt AI works
Open a project in the browser with nothing to install, install a native build for offline work, or read how the pieces fit together first.
New here? The homepage walks through the annotation tools, the dataset features and the export formats.Back to the homepage