Skip to content

Annota Studio · Models

Your data. Your models.

Build AI around the images, documents, and expertise that make your business unique. Start from reviewed data: object detection is in private beta, and more model types are coming soon.

Request model accessExplore model types

Train object detection on your reviewed images. Private beta.

Model types

Choose your starting point.

Image datasets and document datasets open different paths to custom AI.

Images

Models that see your world.

Teach a model to recognize your products, components, and defects from the images and labels your team has prepared.

  • Object detection

    Private beta

    Train a YOLO model to recognize and locate objects in new images using bounding boxes.

  • Instance segmentation

    Coming soon

    Train models to identify the precise outline of each individual object.

  • Image classification

    Coming soon

    Train models to assign a category to an entire image.

Prepare an image dataset

Documents

Models that learn from your documents.

Bring your terminology and domain-specific examples into a language model. Reviewed document datasets are the foundation for this workflow.

  • Language model (LLM)

    Coming soon

    Fine-tune a language model with reviewed examples derived from your documents.

Start preparing your source material today with OCR, text correction, and dataset versions.

Prepare a document dataset

Inside the training workflow

From reviewed data to a model you can use.

With object detection training, you prepare data, evaluate results, and download the model in Annota Studio.

  1. Choose your dataset version

    Start from a saved version of your images, so every run uses approved data and a stable reference.

  2. Configure and train

    Set your training options and follow the progress of each run.

  3. Evaluate the results

    Inspect precision, recall, and mAP50-95 for each run, and duplicate a configuration to try a change.

  4. Take your model with you

    Download the model weights for your own workflow, and keep preparing data for the next round.

Frequently asked questions

How do I get access?

Send us your use case through the model access form. Access is enabled for your organization after approval, independently of your subscription plan.

Do models use storage?

Yes. Retained model archives and generated files count toward your organization's storage allowance, as do the dataset versions used to create them. Training needs available storage for its output.

Does training use credits?

Object detection training currently uses no credits during the private beta. Access requires separate approval, on any plan. AI annotation, OCR, and audio transcription in Studio and Brain use your organization's credits (Brain chat uses each member's weekly allowance first). The Pricing page lists their rates.

How do documents become model training data?

LLM fine-tuning is coming soon. It will start from extracted text that you have reviewed and corrected, using examples derived from that material.

How does this relate to LLM Wiki and Brain?

Models is the part of Annota Studio where you train custom AI models. LLM Wiki exports source documents and instructions for an external AI agent. Annota Brain answers questions from knowledge bases built on your document datasets, with verified sources.

What will your model do?

Tell us about your images or documents and the result you want to achieve. We'll help you identify the right path.

Request model access