# Gopro AI vision test
## Overview
This prototype demonstrates a real-time AI vision system using a GoPro Hero 3, an ATEM switcher, and a computer running object detection via [YOLOv8](https://yolov8.com/). The project showcases how machine learning can be integrated into a live production workflow to identify and label objects on a tabletop workspace.

## Goals
- Prototype a functional "smart surface" that visually identifies tools and objects in real time.
- Integrate AI detection into existing AV infrastructure via SDI and BlackMagic DeckLink.
- Provide a demonstration of AI/ML capabilities for iHub fabriation use.

## Hardware Setup
- **Camera:** GoPro Hero 3 (Mini HDMI out)
- **Signal Chain:**
- GoPro → HDMI to SDI Converter → ATEM Input
- ATEM SDI Out → DeckLink SDI In (Computer)
- Computer HDMI Out → HDMI to SDI Converter → ATEM Input (Processed Feed)

## Software
- **AI Detection:** YOLOv8 (Ultralytics) running in Python
- **Vision Processing:** OpenCV
- **Routing Tools:** OBS Studio for ingest/output management with DeckLink I/O

## Outcome
The system creates a live, labeled video feed that visually highlights objects placed under the overhead camera. It demonstrates practical AI integration in a maker/learning space and can be expanded to include dataset training, Slack alerts, or audio feedback.

## Future Ideas
- Custom object training for fabrication tools
- Integration with UDL/accessibility tools
- Hands-on student workshops to build similar systems