Image labeling FAQ for plans, AI pre-labels and data
This image labeling FAQ answers the questions ML engineers and their managers ask before they move a dataset into ImageLabeling. It covers how the auto labeling works, which formats you can export, what happens to your images, how plans and the AI quota work, and how data annotation services fit in. If something is missing, write to [email protected].
Image labeling FAQ about the product
ImageLabeling is a browser based workspace for labeling images for computer vision. These answers cover what it does and what it does not do.
What does ImageLabeling do?
You upload images, an AI model draws the first boxes, polygons, masks, keypoints or image classes for your label schema, your team accepts or fixes each object in a keyboard first editor, a reviewer approves the work, and you export a versioned dataset in COCO JSON, YOLO txt with data.yaml, Pascal VOC XML or CSV. See how it works for the full workflow.
Which annotation types are supported?
Starter covers bounding boxes, polygons, points and image classes. Team adds segmentation masks and attributes per object. Business adds keypoint skeletons and custom schemas. Enterprise includes every annotation type. The data annotation tools page explains each tool and its shortcut.
Which formats can I export?
Every plan exports COCO JSON, YOLO (one txt file per image plus data.yaml), Pascal VOC XML and CSV. Exports are built from a frozen dataset version, so the files you train on match the labels your reviewers approved. The YOLO labeling tool and COCO annotation tool pages show the file layout.
Can I import labels I already have?
Yes. You can import existing labels in COCO, YOLO, Pascal VOC and CSV. Imported labels appear in the editor like any other label, so your team can review, fix and version them before the next export.
Can several people label at the same time?
Yes. Each plan includes seats: 3 on Starter, 10 on Team, 30 on Business and unlimited seats on Enterprise. Roles are owner, admin, reviewer, labeler and viewer, and you can limit access per project.
Does ImageLabeling work with video?
You can upload video frames as images and label them like any other image. A native video timeline with frame interpolation is not part of the product.
Do I need to install anything?
No. ImageLabeling runs in the browser, so you can annotate images online from any modern desktop browser. Teams that want automation use the API on Team and higher plans.
Questions about AI pre-labels
AI pre-labels are a first draft that saves drawing time. A person stays in charge of every label that reaches your export.
How accurate are the AI pre-labels?
They are a first draft. Every object comes with a confidence score, so your team can sort by confidence and check the uncertain ones first. Reviewers accept, edit or reject every object, and when review is on, nothing goes to export unreviewed. Accuracy depends on your images and classes, so we do not promise a fixed number.
What counts toward the monthly AI image limit?
One image pre-labeled by AI once counts as one image. Manual labeling, edits, reviews and re-exports do not count. The limits are 1,000 images per month on Starter, 3,500 on Team, 12,000 on Business and 35,000 on Enterprise.
Does deleting images free up AI quota?
No. Deleting images never frees the monthly AI quota, because the AI work on those images was already done. The quota resets with your next billing period.
Can the AI label my own custom classes?
Yes. The model labels for the classes in your label schema, and you can add a class prompt such as "helmet, person, forklift". Business and Enterprise also support custom schemas with hierarchical classes and attributes.
Data and security questions
Your images stay yours. The answers below summarize the data security for image labeling page.
Is my data used to train AI models?
No. Your images are processed only to produce labels for your dataset. They are never used to train third party models.
How is my data protected?
Data is encrypted in transit and at rest. Access follows roles and project permissions, Business adds Google and Microsoft SSO and an audit log, and Enterprise adds SAML SSO and a data processing agreement.
Can I delete my data?
Yes. You can delete images and projects in the workspace, and you can ask [email protected] to delete your whole workspace. Images uploaded to the homepage demo are deleted after 24 hours.
Who owns the images and the labels?
You do. You keep all rights to the images you upload and to the labels your team creates, and you can export them in open formats at any time.
Plans and billing questions
All plans are subscriptions, billed monthly or yearly. Yearly billing saves 50 percent and is selected by default on image annotation tool pricing.
Can I try ImageLabeling before I pay?
Yes. A free demo on the homepage labels one image, so you can see the AI pre-labels and the editor on your own picture. Plans start at 24 USD per month billed yearly.
How much does ImageLabeling cost?
Starter is 49 USD per month, or 24 USD per month billed yearly (288 USD per year). Team is 149 USD, or 74 USD billed yearly (888 USD per year). Business is 499 USD, or 249 USD billed yearly (2,988 USD per year). Enterprise is 1,499 USD, or 749 USD billed yearly (8,988 USD per year).
Can I change plans later?
Yes. When you move to a higher plan or from monthly to yearly billing, the difference is charged right away and the new limits apply once the payment goes through. Moving to a lower plan takes effect without a refund of the current period.
Can I cancel anytime?
Yes. Cancelling stops the next renewal and you keep access until the end of the period you paid for. Payments are final, as described in the refund policy.
Do you send invoices?
Every payment comes with a receipt. Enterprise includes invoicing for teams whose finance department needs it.
Managed labeling questions
Some teams want people as well as software. Managed labeling adds vetted annotators to your workspace.
Do you offer people to do the labeling?
Yes. On Business and Enterprise you can add vetted annotators as a monthly retainer, from 899 USD per month per dedicated annotator. They work inside your workspace under your label schema and QA rules. Read more on data annotation services.
How do you keep annotator quality high?
Annotators work in the same labeler then reviewer workflow as your own team. Reviewers approve or send work back with a comment, consensus runs on a sample of images, and Business shows a QA score per labeler.
Can I join as an annotator?
Yes. Experienced annotators can apply to the vetted workforce on the annotators page.
Label your first image now
Upload one image to the free demo and see the boxes, masks and points drawn for you. When you are ready, pick the plan that fits your team.