Quickstart

Datature Vi is an end-to-end VLMOps platform that helps you transform raw data into a trained VLM and deploy it quickly. Whether you're building defect detection systems, product classifiers, or custom vision AI solutions, Vi streamlines the entire workflow—from data preparation to deployment.

This quickstart guide walks you through the complete process using your own data. You'll learn how to prepare your dataset, train a VLM, and deploy it for real-world use.

⏱️ Time to complete: ~30-45 minutes

📚 What you'll learn: Complete VLM workflow from data to deployment

📋

Prerequisites

• A Datature Vi account (free sign-up)
• Images and annotations ready for upload
• Understanding of VLM concepts (optional but helpful)
• Familiarity with phrase grounding or VQA

Sign up for free →

📘

New to Datature Vi?

You'll need an account to get started. Sign up for free!

💡

Want to understand the concepts first?

Learn about Phrase Grounding and Visual Question Answering before diving in.

Next steps

Follow these three steps to get your VLM up and running:

Related resources

• Prepare your dataset — Upload images and create annotations
• Train a model — Configure training and create workflows
• Deploy and test — Download models and run inference
• Concepts — Learn about VLM concepts and terminology
• Phrase grounding — Understand object localization tasks
• Visual question answering — Learn about VQA capabilities
• Create a dataset — Detailed guide to dataset creation
• Annotate data — Use the annotation tools
• Vi SDK — Programmatic access using Python SDK
• Evaluate a model — Assess model performance
• Manage models — Rename, download, and organize models
• Contact us — Get help from the Datature team