Skip to main content

AI software can identify objects in photos and videos at near-human levels

A new AI software program developed by researchers at Google and Stanford University can recognise objects in photos and videos at near-human levels of understanding.

ai software program google stanford university object recognition technology images videos human level understanding

It was only recently that computer systems became smart enough to identify unknown objects in photographs. Even then, it has generally been limited to individual objects. Now, two separate teams of researchers at Google and Stanford University have created software able to describe entire scenes. This could lead to much better and more intelligent algorithms in the future.
Stanford's work, entitled "Deep Visual-Semantic Alignments for Generating Image Descriptions", explains how specific details found in photographs and videos can be translated into written text. Google's version of the technology, in a study titled "Show and Tell: A Neural Image Caption Generator", produced similar results.
While each team used a slightly different approach, they both combined deep convolutional neural networks with recurrent neural networks that excel at text analysis and natural language processing. The programs were able to "learn" from each new interaction, with algorithms enabling the system to improve its accuracy by scanning scene after scene, looking for patterns, and then using the accumulation of previously described scenes to extrapolate what is being depicted in the next unknown image.

ai image recognition

"The system can analyse an unknown image and explain it in words and phrases that make sense," says Fei-Fei Li, a professor of computer science and director of the Stanford Artificial Intelligence Lab. "This is an important milestone. It's the first time we've had a computer vision system that could tell a basic story about an unknown image by identifying discrete objects and also putting them into some context."
These latest algorithms are being trained on a visual dictionary – the ImageNet project – with a database of more than 14 million objects. Each object is described by a mathematical term, or vector, that enables the machine to recognise the shape the next time it is encountered. Those mathematical definitions are linked to the words humans would use to describe the objects.
“I was amazed that even with the small amount of training data that we were able to do so well,” said Oriol Vinyals, a Google computer scientist who worked with members of the Google Brain project. “The field is just starting, and we will see a lot of increases.”
In the near term, computer vision systems that can discern the story in a picture will enable people to search photo or video archives and find highly specific images. Eventually, these advances will lead to robotic systems able to navigate unknown situations. Driverless cars would also be made safer. However, it also raises the prospect of even greater levels of government surveillance.

 frisbee 
"A group of young people playing a game of Frisbee."
 

 frisbee 
"A person riding a motorcycle on a dirt road."
 

 frisbee 
"A pizza sitting on top of a pan on top of a stove."
 

Comments

Popular posts from this blog

Google Announces Android Wear Update With WiFi Support, Always-On Apps, And More

It has been a while since Android Wear got any substantial updates, but today Google is announcing a big one. A new version of Wear will be rolling out over the coming weeks that includes a number of previously rumored features (like WiFi support) and some all new stuff (like always-on apps). Most Wear devices use the always-on ambient mode for the watch face by default, the Moto 360 being a notable exception. The new Android Wear version allows apps to operate in ambient mode too, so they remain active when the watch goes to sleep. That makes it easier to take a quick glance at the app instead of waking the device up and opening the app all over again. The watch will still only go into full-color mode when necessary. WiFi support is also coming in the update, which means your watch can be useful even if your phone isn't connected. Watches with WiFi support will be able to connect to WiFi and still get messages and notifications from your phone, provided it has an interne...

Budding #entrepreneur from Chandigarh University!!

Budding #entrepreneur from Chandigarh University!! #CU #students unfolded their creative ideas and presented them with a productive shape! Meet Our #Automobile #Engineering student - Trilok Singh, who has started his own start-up with the name GEARR TECHNOLOGIES under the guidance of CU-TBI. This start up focuses on affordable high end #Bicycles and its high #technology equipment’s. This start- up will bring to the Indian audience the scope of Products, #innovation, creativity and customization available in the market. Watch the video!!

South Korea aims for startup gold

Back in 2011, when South Korea won its longshot bid to host the 2018 Winter Olympics, the country wasn’t widely recognized as a destination for ski and snow lovers. It wasn’t considered much of a tech startup hub either. Fast forward seven years and a lot has changed. For the next 10 days, the eyes of the world will be on the snowy slopes of PyeongChang. Meanwhile, a couple of hours away in Seoul, a burgeoning startup scene is seeing investments multiply, generating exits and even creating a unicorn or two. While South Korea doesn’t get a perfect score as a startup innovation hub, it has established itself as a serious contender. More than half a billion dollars annually has gone to seed through late-stage funding rounds for the past few years. During that time, at least two companies, e-commerce company Coupang and mobile-focused content and commerce company Yello Mobile, have established multi-billion-dollar valuations. To provide a broader picture of how South Korea stacks ...

The Eight Most Impactful Excel Shortcuts That You Should Master

If you’ve ever gone online to research improving your Excel skills, you’ve undoubtedly come across a post or two listing all of Excel’s keyboard shortcuts.  In the latest version of Excel, Microsoft has made it easier than ever to learn shortcuts, by assigning shortcuts to nearly every function and making the discovery of the input sequence very transparent. While memorizing Excel shortcuts will generally improve your productivity, not all shortcuts are created equal.  Shortcuts that you never use are not inherently not very useful and not worth memorizing.  Your focus should be on the shortcuts that have the most impact – either by the amount of time it saves you, the frequency that you’ll use them, or the behavior it encourages. If you’ve already started using Excel or just haven’t utilized shortcuts heavily before, review the top eight shortcuts below.  For any that you don’t know already, I would suggest memorizing them and incorp...

Facebook will verify the location of U.S. election ad buyers by mailing them postcards

Facebook’s global director of policy programs says it will start sending postcards by snail mail to verify buyers of ads related to United States elections. Katie Harbath, who described the plan at a conference held by the National Association of Secretaries of State this weekend, didn’t reveal when the program will start, but told Reuters that it would be before the Congressional midterm elections in November. The cards will be sent to people who want to purchase ads that mention candidates running for federal offices, but not issue-based political ads, Harbath said, and contain a code that buyers need to enter to verify that they are in the U.S. The program is similar to ones used by Google My Business and Nextdoor when they need to verify business owners or users who want to join closed neighborhood groups, respectively. Harbath told Reuters that the postcards “won’t solve everything,” but were the most effective method the company came up with to prevent people from using fa...