If you have a Kinect you can join in a new project to map the entire world in 3D - well at least a as much of it as possible. Kinect@Home is a community effort to input 3D models to a server.
All you have to do is download some drivers and a plugin and use your Kinect to scan the area you want to make a 3D model of. The data is uploaded and your reward is a 3D model that you can embed in a web page. You can also download the model to use in your own software.
The following video explains how it all works.
What is the point of all of this?
The simple answer is robots need "big data". It may be a catchphrase, but when it comes to machine learning the more data the better. By establishing a library of 3D models, robots and machine learning software can improve their recognition and planning.
You can see a sample of the 3D model embedded in a browser below:
There is a gallery of models that have already been recorded.
An even more ambitous project would be to provide robots with a complete 3D map of an environment. This would make it possible for it to work out where it was in the real world and move to a new location. Of course, it could feed any new features it found back into the map. At the moment Kinect@home doesn't record GPS positions, so the models are purely local.
Yes, another neural network breakthrough. This one provides us with a way of training using reinforcement learning, but without the need for the biologically implausible and expensive back propagation [ ... ]
A new, open-source time series database built with the Postgres engine has been released. TimeScaleDB is currently available in a single-node version, and is optimized for fast ingest and complex [ ... ]