$16K Cameras and Hard Lessons: How We Built an Advanced AI Vision Pipeline (and Rebuilt It, Twice)

"...the book for the online entrepreneur" Designing Success SEO Tactics for AI Driven Campaigns





Prompt Engineering The Handbook: Practical Applications of Prompt Engineering Kindle Edition



1 long video, 10 viral clips. Create 10x faster.

OpusClip turns long videos into shorts, and publishes them to all social platforms in one click.

149 Views

  • Thanks! Share it with your friends!

    You disliked this video. Thanks for the feedback!

philigaha
Published
Building an AI vision pipeline for real-world robotics comes with unique challenges. At Urban Machine, we learned hard lessons moving from consumer-grade cameras to high-end structured light 3d systems, and finally to a scalable, cost-efficient setup using 2d Lucid cameras.

In this video, I dive into:

- Why heuristic CV fails for unpredictable materials like wood.
- How continuous material flow transformed our pipeline.
- Why pipelines, not models, are the key to scaling AI systems.

If you’re working on AI vision or robotics, this breakdown of our approach—and what we’d do differently—might save you some headaches.

Check out some open source tools we built that helped make this possible:
- our ros tools: https://github.com/urbanmachine/node_helpers
- 3d robot tools: https://github.com/UrbanMachine/onshape-urdf-exporter
- containerized ros tooling: https://github.com/UrbanMachine/create-ros-app

Reach out via alex@urbanmachine.build
Category
Types of Artificial Intelligence
Be the first to comment
PC CHIPS UK