... this is how we would control our drones in the near future, if this promising object recognition technology from Google goes mainstream (and gets an API). I think drones and other mobile robots could benefit a lot. Check this out:
Automatic object recognition in images is currently tricky. Even if a computer has the help of smart algorithms and human assistants, it may not catch everything in a given scene. Google might change that soon, though; it just detailed a new detection system that can easily spot lots of objects in a scene, even if they're partly obscured. The key is a neural network that can rapidly refine the criteria it's looking for without requiring a lot of extra computing power. The result is a far deeper scanning system that can both identify more objects and make better guesses -- it can spot tons of items in a living room, including (according to Google's odd example) a flying cat. The technology is still young, but the internet giant sees its recognition breakthrough helping everything from image searches through to self-driving cars. Don't be surprised if it gets much easier to look for things online using only vaguest of terms.
more to read:
source: http://www.engadget.com/2014/09/08/google-details-object-recognition-tech/?ncid=rss_truncated
source: http://googleresearch.blogspot.com/2014/09/building-deeper-understanding-of-images.html
Comments
Thinking about self-driving cars, some of my friends think that a significant test will be self-navigation to safely park a car or truck on a ferry. See for example the following two images. There is basic training for ferry workers, however the actual behavior can vary by individual.
This is really a great achievement !
This may change ;)
On the internet, nobody knows you're a dog.
Yeah but i guess Google made some improvements. It may work better than openCV...
I would suggest to couple a mathematical morphology algorithm as an additional input to the neural network. I have shown in 1993 during my thesis that it made possible with the computers of that time (less powerful than today's phones) to recognize objects in a few seconds with a reliability of 90%, with very limited neurones in the net. I should dig this thesis out of my archives someday...