SensAI Kits 🎎

This is a collection of context-aware AI kits, fully integrated with the Meta XR SDK and Unity.
Vision kits are built to support Meta Camera Access (PCA).

📝 Table of Contents

1. AI Capabilities Workbench
2. Zero-Shot Object Detection Kit
3. Custom AI Model Training Kit (Roboflow)
4. Poker AI Assistant (Roboflow)
5. Relevant Sources & Opportunities
6. License

Overview

1. AI Capabilities Workbench

🎯 A Unity based workbench implementing 7 workflows and 10 providers, to bring contextual AI to your apps. Compatible with Meta Quest.

• Multi-AI Provider Support – Supports Nvidia, OpenAI, Google Gemini, Groq, Roboflow, Stability AI, and AWS
• Pre-configured scenes with easy API key management, enabling quick AI-powered XR prototypes

⚠️ Limitations

• Requires API Keys: Users must register with the AI providers and configure their own API keys for detection to work
• Requires Internet: Some of the workflows send calls to cloud API

GitHub: 👉 AI Capabilities Workbench

---

2. Zero-Shot Object Detection Kit

🎯 A Unity plugin that enables real-world object detection in XR using Microsoft Florence-2 on Meta Quest.

• Instantly detect objects in your environment with zero-shot AI (no training required)
• Send image data from your Quest to the Florence API and receive rich detection & description results
• Fully integrated with Unity for easy setup and flexible use in XR workflows

⚠️ Limitations

• Not Real-Time: API response times mean detection is fast but not instantaneous
• Requires Internet: Wi-Fi needed to send images to the cloud API
• No Timestamps: Bounding boxes are not time-synced

GitHub: 👉 Zero-Shot Object Detection Kit

---

3. Custom AI Model Training Kit (Roboflow)

🤖 A Unity plugin that brings you custom-trained object detection to XR — powered by Roboflow and optimized for Meta Quest.

• Upload and annotate your own image datasets (capturing directly on Meta Quest recommended)
• Run models locally on-device for faster inference and offline use
• Fully integrate detection results into your Unity XR app

⚠️ Limitations

• Setup Time: Requires effort to collect, annotate, and train datasets
• CUDA & Docker Setup Needed: See Roboflow’s repo

GitHub: 👉 Custom AI Model Training Kit

---

4. Poker AI Assistant (Roboflow)

♣️ A Meta Quest application that uses computer vision and Poker Odds API to calculate poker hand strenght in real time.

• Detects poker cards and calculates odds and win probabilities
• Runs 100,000+ simulations for highly accurate predictions using the Poker Odds API
• Performs local inference directly on Meta Quest
• Minimalistic UI/UX optimized for wearables

⚠️ Setup Notes

• Wi-Fi Configuration: Ensure Meta Quest is on the same Wi-Fi as your server
• Server Setup: Local Roboflow inference server and Node.js Poker Odds server must be running
• Permissions & IP: Verify permissions on Meta Quest and correct IP address in RoboflowCaller
• CUDA & Docker Setup Needed: See Roboflow’s repo

GitHub: 👉 Poker AI Assistant

---

5. Relevant Sources & Opportunities

• SensAI Hackademy - Early access program for courses and toolkits
• SensAI Hack - Upcoming hackathons where you can use the kits

---

6. License

📜 By downloading and using these kits, you agree to the License Terms.