Average Joe

Look Mum! No wires!

Traveling shows need a system that can be installed quickly and be ready to go wherever the event takes place. We’re designing Octopi to set up in under 15 minutes.

Our low-latency pipeline processes footage at the edge, without sending every camera feed back to a remote workstation. That’s what opens the door to a truly wireless setup, with less cabling across the venue.

We’ve also built in a neat little battery pack from @RSDistributor, giving us an estimated 2–3 hours of 5MP capture at 120 fps.

The hardware is taking shape, with camera technology from @Basler_AG and edge computing from @nvidia . Arrive, install, and go.

#LiveEvents #ComputerVision #EdgeAI #SportsTechnology #Octopi

I’d planned to spend the first half of this week getting the camera powered up and running through the Jetson.

Twenty minutes later, it was working.

In typical Jetson and Basler fashion, the bit I’d set aside days for took less time than making a cup of tea. So now I’m re-planning the next couple of days around redesigning our janky Frankenstein of a pipeline 😂

A very good problem to have.

Thanks again to @Basler_AG and @nvidia for making it an easy ride so far.

One camera. One feed. The bigger picture is coming into focus.

#ComputerVision #MachineVision #EdgeAI #Robotics

On paper this does about 90% of my custom multi-camera pipeline in one model 👀

Has anyone actually used MVMP-HMR?

Is this concept/research-stage or something that’s usable in reality? Essentially a post-processing pipeline, or is there a credible path to real-time inference?

Very interested in talking to anyone working with it.

#ComputerVision #HumanPoseEstimation #HMR #MultiView

MVMP-HMR paper
arxiv.org/pdf/2607.24302

The first bit of Octopi hardware is on the bench 🐙

Very excited to start building on NVIDIA Jetson.

8 cameras. 8 edge processors. One real-time understanding of what’s happening in the space.

Time to make it see.

@kGolid I hope you like this one too!