DeepLens AI Data Mining

https://aws.amazon.com/deeplens/

I have a particular interest in High Res cameras and application of Visual AI to what they are capable of learning.  There are many entries in this blog related to them.  This is the latest.  A $250 High Res (don't know how high yet) camera with AI.  Not so new but perhaps a leap ahead of what cameras available now can do?

This camera becomes a sensor device on the IoT to stream data back to AWS where the heavy lifting of Deep Learning (anything desired to be learned by person or machine)  takes place.

Res 1080p video  4 megapixel camera.  Not that impressive compared to resolution that can identify a face on the other side of a football field in a facial recognition program.  There is however a lot of low hanging lower resolution fruit out there to be harvested with this camera. The AI probably works just as well on an extremely Hi-Res gigapixel level video or picture.

It is not just a camera but an AI system.

Maybe fast enough and with sufficient resolution to be a poor man's LPR reader?  Put it on a car and feed license plate numbers to an aggregator as "proof of work" for an LPR Coin at the rate of a fraction of a real cent (like a 1 cent mil value) per plate for mining license plate ID time and location.

That is monetizing a cross country truck driver's data base collected with a cheap camera.

Business Model:  Monetize the Data Base.

Don't have one to monetize?

Create one on the cheap to output aggregate value from low level single element input collection and then sell it (stream it) to a buyer.

It is a new Gig job!

Side money for Uber drivers?

Conceptually, with enough of these camera on vehicles in a few years earning a stream of income they could, aggregating their streaming input keyed to location, create in a day what took Google years to do with a special purpose camera mounted on a large number of cars driving all the streets in the USA.  Each individual camera pointed in a different direction.  The AI program the chooses the best clearest picture of the same location from an X period of time collection of input and create the visual street map.

That is how fast technology is moving.  It puts the Google camera and technology in a museum.

Proof of concept:  There was an interesting trial of taking all the Google Image pictures of a major point of interest (Eiffel Tower) and compositing them in a 3 dimensional navigable single picture.   It is just stitching done by AI and picture recognition.