Artificial Intelligence and Robotics blog
Awesome-o
This user hasn't shared any biographical information
Posts by Awesome-o
Google goggles for the Android
Dec 8th
Earlier today, Google Labs released the Google Goggles application for their Android operating system, i.e., for running on Android-powered smart-phones. Goggles allows users to search the Web using images as input; it allows users to perform a visual search with little effort. Once the application has been downloaded and installed taking a photo with the phone’s camera allows users to instantly obtain information about the imaged object via the application automatically performing a search online using the traditional Google search engine.
Users can take photos of books, CDs, and artwork to get more info about pricing, e.g., on Amazon, or additional information about the author, musician or artist. The application also comes equipped with Optical Character Recognition which means that it can automatically extract information from a business card and add it to one’s personal contacts. Moreover, Goggles can recognize landmarks and using a phone’s GPS and compass augment a live video with information about local businesses that are in view.
I don’t know how well the application works but it looks awesome in the promotional video below. Considering that Google Goggles is still a Labs application, I suspect that even if it doesn’t work perfectly now it won’t be long before it does. I wonder if Google will also port Goggles to the iPhone(see also a similar but more restrictive application for the iPhone called Nearest Tube.)
[source Google]
A biologically-inspired vision system
Dec 6th
Researchers at the Rowland Institute, Harvard, and McGovern Institute for Brain Sciences, MIT, are developing new, biologically-inspired vision systems taking advantage of faster computers. Their goal is to create vision systems for image understanding that can be as accurate as biological systems and more specifically the human visual system. The researchers have developed a new method that allows them to evaluate many different vision systems and quickly determine which are best suited for scene understanding. In a PLoS Computational Biology paper, the researchers show that their method performs better than current state-of-the-art computer vision systems when tested using standard data sets.
If you don’t want to read the paper, then you should at least watch the below video in its entirety. In the video, lead researcher David Cox explains at a high level how biological vision works and how their computational system mimics it to achieve the results presented in the paper.
Intel’s 48-Core Single-Chip Cloud Computer
Dec 3rd
Intel announced yesterday a new 48-core single chip computer that is low power and promises huge speed increases for applications that can take full advantage of parallel computing. Applications in Artificial Intelligence including machine learning, robotics and computer vision stand to greatly benefit from such silicon.
If you remember, a bit less than 3 years ago, I wrote about Intel’s Teraflop Research Chip that was promising a 80-core design within 5 years. I wrote then,
Intel executives speculated that the new chip will be available for desktop and laptop PC use in 5 years although the most conservative ones predict that what is most likely to be available in this time frame is a smaller chip with maybe 20 to 40 cores.
Intel practically delivered on that promise yesterday when they announced their new 48-core chip. The new architecture brings cloud computing to a single chip which means that to a large degree the new microprocessor is going to be aimed at server applications. However, we all know that eventually such a technology will find its way into our desktop and laptop computers. In fact, in the Press Release, Intel notes how the new chip could be used to develop new and exciting consumer-based applications,
For example, future laptops with processing capability of this magnitude could have “vision” in the same way a human can see objects and motion as it happens and with high accuracy.
Imagine, for example, someday interacting with a computer for a virtual dance lesson or on-line shopping that uses a future laptop’s 3-D camera and display to show you a “mirror” of yourself wearing the clothes you are interested in. Twirl and turn and watch how the fabric drapes and how the color complements your skin tone.
These are fascinating news. So far, NVIDIA was leading the way with their GPU-based parallel processing solutions (see for example the Tesla desktop supercomputer) but now Intel is back in the game. The competition between these two companies will drive innovation and so I expect to see chips with even more cores in the not so distant future.
Along with the Press Release, Intel also released the below video outlining the features in their new 48-core single-chip cloud computer.
Zhu Zhu Pets: Robot gift for someone special this holiday season
Dec 1st
While this will not go down as one of the great innovations of science, it comes as little surprise that one of the most popular and financially successful type of robots is determined so by the kid-to-preteen demographic. We’re talking about the Zhu Zhu Pets
They are not the most sophisticated piece of machinery in the world, but can do a number of tricks that little kids love. They react to physical interaction, make random hamster noises, explore spaces and rooms on their own, they can run on a hamster wheel and come with several other accessories. Basically they do the same things as a real hamster, but without the smell or the need to clean up after them. The fact that they run on batteries also makes them more convenient, and parents do not have to worry about them dying in a couple of years or sooner.
UGOBE had a chance to be the first on the market with such a cool toy years ago but their over hyped Pleo dinosaur robot failed for being too expensive and not delivering on the company’s promises. Zhu Zhu pets is a perfect example of how to make a simple robot toy that kids would love.
Mechanical toys as “pets” is not something new (there are many more out there other than Pleo,) but the greatly successful formula about these robots is that they hit that sweet spot between quality and affordability. It also suggests that if the alternative is realistic and cheap enough, many people would prefer it over a living animal. The problem with robotic pets in the past was either that they were too simple and did not fulfill the desires of the consumers, or they were way too expensive and people would rather just buy the real thing. But give people what they want at the price they want, and they will buy out the shelves.
Does this piece of news endanger the family dog or cat? Maybe not for now, but in the future Rover and Belle might definitely be looking over their shoulder.
[source Zhu Zhu Pets]
Retinal implant helps blind man see again
Nov 29th
This story sounds like it is taken out of a science fiction novel but news reports around the world say that Mr. Peter Lane, 51, from Stretford, has his lost sight partially restored using a new device under development by California-based Second Sight Medical Products Inc.
The story is that Lane is one of several blind people who are testing the new device known as Argus II Retinal Stimulation System and he is the one who recently found fame in the media having his lost sight partially restored to the extend that he can now read some small words and navigate the outside world with increased ease.
According to Second Sight, the device consists of a small camera and an electrode-studded array implanted on the patient’s retina. This is how it all works.
The camera on the glasses captures an image and sends the information to the video processor, which converts the image to an electronic signal and sends it to the transmitter on the sunglasses. The implanted receiver wirelessly receives this data and sends the signals through a tiny cable to the electrode array, stimulating it to emit electrical pulses. The pulses induce responses in the retina that travel through the optic nerve to the brain, which perceives patterns of light and dark spots corresponding to the electrodes stimulated. Patients learn to interpret the visual patterns produced into meaningful images.
The Argus 16 and the next generation Argus II (which has 64 electrodes) are not designed to give sight to those who never had it. Mr. Lane and the others in the clinical trial suffer from Retinitis Pigmentosa which means they lost their sight in their later years. This is important because it essentially means that the brain is capable of interpreting the signals received from the eye but the disease has destroyed the retina and the flow of data has seized. Argus replaces the damaged tissue with an electronic one allowing the brain to once again receive the signals necessary to perceive the world.
This is a great breakthrough in the fight against blindness. Statistics show that more than 200,000 people in the USA and Europe suffer from Retinitis Pigmentosa and Argus II can immensely increase quality of life for these people. Some suggest that Second Sight might be able to start selling the device commercially in one year at the cost of nearly $100,000. Sure it is expensive but it is a start. Eventually, when the device is proven to work well and the cost of production decreases it will become much more affordable and available to all who need it.
[For clinical trial information go here.]
[source Times Online and the Daily Mail.]
Cyborg Beetles
Nov 26th
The Pentagon have been dabbling in the field of robotics for some time now, and one of their newer projects is buzzing nearer completion. We’re talking about remote-controlled beetles that can be stirred in the right direction, but allowed to fly on their own. In the future, plans are to equip these insects with tiny cameras to help search for survivors of disasters, or with sensors to sniff out potential chemical weapons.
Watch the video of a remote-controlled beetle below.
To make it clear—these beetles are not robots, but cyborgs. They are actually living creatures that have been implemented with electronic parts to influence their movements, not unlike reigns on a horse. To help them adjust, this “modification” takes place in the pupa stage when they are still developing. When they are fully grown, the beetles can be commanded by electric signals controlled by a laptop that are transmitted to their nervous systems. Researchers at the Agiltron Corporation, who are contracted to the project, plan to implement another system in the insects that will have them twitch when they detect hazardous chemicals.
Other, more vocal insets, like crickets, are also lined up for modification. They will be equipped with MEMS transceivers that will receive specific sounds made by other cyborg crickets, formulate the data and carry it on to other insects. This will create a sort of wireless network of information that can be remotely monitored.
This is, of course, a very interesting development, albeit a bit reminiscent of sci-fi horror movies. Some have questioned the ethics of the procedures, and wonder if or where a line will be drawn. If we start out with insects, would we move on to hamsters, rabbits, dogs…one day even illegal human experiments?
Nonetheless, one of the bigger problems right now is actually getting the insects to live longer. The modifications take their toll, and the beetles burn out before they can be used for something other than research.
[source NewScientist]
Space robotics: One Giant leap for Robots
Nov 25th
Robots have been employed by a number of NASA’s missions in the past, particularly those to the Moon, where they have been used to scan the surface and collect samples and various kinds of data. NASA has also been a pioneer in semi-autonomous robots exploring the surface of Mars; the twin Mars rovers Spirit and Opportunity are still operating on the surface of the Red planet several years after their initial mission was completed. The Phoenix robot successfully landed on Mars last year and completed its 3-month mission. Moreover, NASA is preparing to launch soon a new generation Mars rover aka the Mars Science Laboratory while testing the ATHLETE rover for a mission to the Moon.
But none of these very cool robots have been able to walk—or build Japanese rock gardens.
In a presentation entitled “Realization of Moon Exploration Using Advanced Robots by 2020,” Toyota shared plans about the future use of its innovative Toyota Partner Robot series. Initially, these humanoid robots were designed to help with life down here on Earth, specifically with elderly care and assistance. A result of Japan’s aging population and a lack of labor force, they are able to perform a variety of tasks, including transporting passengers and helping citizens in need. Their application, however, is set to be taken quite a big step further—to the moon.
There, the Toyota Robots will assemble space tents and relay stations, as well as deploy solar array and use a telescope. Additionally, they will work alongside a Rover to collect further data of the moon. The presentation also showed them building a traditional Japanese rock garden, which is rather cooler than planting a simple flag. To perform these tasks, they will be equipped with a number of big upgrades. As the specs in the yellow table predict, they will have:
- joints that are protected from regolith
- a small capacity solar battery onboard
- their internal status showing on screen on chest
- arms exchangeable for different tasks
- ability to jump with springs in legs
- keep warm during night covered in a metal cloak
That last one paints quite the endearing picture of our robot helpers working all day and camping out beneath the stars at night on a lonely desolate planet. With 2020 being the proposed date, however, we are still a decade away from finding out whether this is yet possible.
[source in Japanese]
The Visual Memex Model: Modeling object relationships for scene understanding
Nov 23rd
Computer vision researchers have for decades been trying to develop algorithms for scene understanding from images and/or video. No doubt, they have made huge progress towards this goal. For example, in the last decade new methods for feature-based object recognition have been developed with a large degree of robustness to scale, viewpoint and illumination changes. Such methods are what makes products such as Photosynth possible today.
However, full scene understanding has continued to elude researchers. Alexei Efros’ group at CMU are now proposing a new method for scene understanding that looks at the individual objects in a scene and their spatial relationships. The Visual Memex Model as they call it is a new method for encoding information about specific objects and their visual similarity and contextual relationships.
The insight behind the Visual Memex model is that the traditional approach in computer vision that objects belong to well defined categories is not correct and that an exemplar-based definition of categories is more suitable. Using evidence from psychology, cognitive neuroscience and other disciplines Efros and his student Tomasz Malisiewicz argue that this new approach is more suitable for scene understanding in computer vision.
In their proposed model, objects are represented by examples of their appearance in images comprising the vertices of a graph in which the edges represent either visual similarity between exemplars or contextual relationships, e.g., a person is often seen next to a car. These relationships are learned automatically from data using state-of-the-art machine learning methods. Given this graph, new objects are first matched to an exemplar and then the contextual relationships are used for scene understanding. An experimental evaluation using a large database of images shows that the proposed method performs better than category-based systems. I suspect that this work may actually cause a small change in thinking within the computer vision community.
The authors will present their work on the Visual Memex Model next month at the 23rd annual Neural Information Processing Systems (NIPS) conference. You can get a copy of the full paper here (pdf).
Mine-detecting Bacteria
Nov 20th
And now for something different!
People often get a bit squeamish when they hear the word “bacteria”, but scientists at the University of Edinburgh, Scotland, have engineered a new brand that may actually help clear mine fields. They employed a method called BioBricking, which manipulates and assembles DNA strands for a specific effect, not unlike a small machine. The bacteria, when mixed with a solution and sprayed on a suspected mine field, reacts to the traces of chemicals that have leaked out from the explosives, and turns a bright green.
Dr Alistair Elfick, a professor from the University involved in the project explained:
This anti-mine sensor is a great example of how innovation in science can be of benefit to wider society. It also demonstrates how new scientific techniques can allow molecules to be designed for a specific purpose.
Land mines are still one of most dangerous hazards affecting many countries, especially developing ones, with almost 20,000 new casualties every year. Since their location is unknown, previous techniques of mine-detecting have proven to be slow, costly and mostly ineffective. This special brand of glowing bacteria, however, might be a huge step forward in fighting the problem and once again shows the great potential of technology. While it is still not ready for widespread use, if tests prove successful it is sure to garner great interest.
Similar experiments have been attempted in the past, with flowers designed to change color when grown on a mine-field, but were never given the opportunity to be used in many places. One can understand people’s skepticism when it comes to planting the earth with radioactive plants or glowing bacteria, and if this method is to prove successful, it requires a very high level of planning and organization. But if it can save thousands of lives a year, it is a risk that needs to be taken.
[source BBC]
Recent Comments