More Kinect RGB-D conversion breakouts

We've decided to make up a small batch of Kinect breakouts to gauge interest.

As we've stock of both the donor Kinects and the necessary components it seems a gamble worth taking.

They'll be on Tindie fairly soon, once the enclosure design is finalised.

ESP32-CAM Backpack

Our ESP32-CAM Backpack is now ready for sale on Tindie.

This is designed specifically for making the popular ESP32-CAM development board easier to use in battery powered projects.

Kinect RGB-D conversion breakout

The original Microsoft Kinect is rather old now, but they are still of interest when building low cost robotics projects as they're a very affordable way to get an RGB-D depth camera.

The downside is that they are most definitely not designed for anything other than their original purpose, needing a chunky 12v supply and being rather bulky.

Other people in the Open Source Community did the groundwork in working out how to strip down Kinects long ago to remove the pose detection processor and reduce it to just the RGB-D depth camera.

We've gone through this process ourselves and made a scratch built power supply and USB connection in the past but decided to develop a small breakout that makes the end result tidier. Not being a priority the boards have languished for months but today we finally put one together.

We're not convinced this is viable as a product but it works and was a fun exercise. We're going to build a few units for our own upcoming robotics projects. With used Kinects being almost throwaway cheap you can afford to add one to a project and just leave it in place, unlike modern machine vision cameras.

Results from the backslang test

Broadly, backslang worked but we had a very stressful time in the initial setup as we couldn't get a GPS fix in the location where the terminal was set up.

In order to maintain database consistency, all backslang nodes need a consistently accurate time, at least within human scale ie. 100ms or so. If one node has good time then the time sync feature of m2mMesh ensures that all other nodes should have a time sync on this scale.

Testing had show that the GPS module in use on the terminal could get a solid time and location fix indoors slowly when cold started but quickly once warm. However the location for the terminal was a building with a metal roof and this completely blocked the GPS cold start.

The practical solution was to place the terminal outside the building on a long lead, but this was disappointing from a usability point of view and we'll be making a V1.1 of the terminal with a hardware real-time-clock.