Welcome

Thank you for checking out Isaac 0.1! This document is designed to help you understand and maximize the potential of our demo environment (full playground coming soon). Throughout this guide, we've highlighted important tips to help you get the most out of the Isaac.

Navigate through the sections using the table of contents. If you need additional support, contact us directly at [email protected] or join the Discord community 🔗. We have dedicated channels for prompting tips, bug reports, and more.

Getting Started

This section will guide you through basic operations to get you up and running quickly.

Core capabilities 🔗

Discover the key capabilities that make Isaac powerful and versatile for your needs.

• Visual Grounding
  • Object detection
  • Object attribution detection
  • Counting
• Visual Question Answering (VQA)
  • Scene captioning
  • OCR

Tips 🔗

Suggestions to improve model performance.

• Input & output selection
  • Adding multiple images as input
  • Choosing response style
• Prompt Engineering
  • Single-target pointing
  • Multi-target pointing
  • Concise responses
  • Accurate rationale
  • In-context examples

Core Capabilities

Discover the key capabilities that make Isaac powerful and versatile for your needs.

Object detection (Grounding)

Identify any range of subjects (objects, humans, animals, etc.)

Object attribute detection (Grounding)

Segment by subject attributes/descriptors (color, action, role, etc)

Counting (Grounding)

Try simple language like “How many X are there?” Consider adding: “Point to each.”

Scene captioning (VQA)

Ask any question and get natural language responses to dynamic scenes

OCR (VQA)

Read text such as signage, labels, and handwriting.