Weekly IT Digest #2: Image Recognition Describes Pictures and Completes Them
April 21st, 2016
The main event of the last week was Facebook’s F8 Conference for developers. During his speech, Mark Zuckerberg has made a number of interesting announcements. In addition, Microsoft has introduced an image recognition and description tool called CaptionBot; AI learned to guess winners of talent shows, and changes in the AppStore will make iOS developers to spend some money on search ads.
A platform for bot creation, 360-degree camera, and several innovations for the web-content creators was presented at the Facebook F8 Conference. From now, Instant Articles function - a function for publishing content on the social network - became accessible to all. After a short patch of their RSS-feed, the content-sites and blog owners will be able to publish their articles directly to Facebook. The social network, in turn, promises to share the income from advertising with them. In addition, Facebook will allow saving articles for later reading directly from the page they are placed on and share not only the entire article but also its individual fragments.
Mobile app developers should pay attention to the upcoming changes in the App Store. Apple is preparing to sell “seats” in the search results of its online store using the same principle Google uses to sell search ads. Obviously, this is going to complicate the mobile application promotion and force developers to invest more in their apps.
Artificial Intelligence algorithm, developed by Alibaba Group, was able to predict the winner of the Chinese song show “I’m a Singer.” Aliyun platform analyzed many parameters including the popularity of the song, singer’s change dynamics, performance’s energy, and even the audience’s reaction. This is Aliyun’s second attempt to guess the winner. The first was a pilot and was tested in the previous season of the show, after which the algorithms of the system were changed. The winner of the show is determined by 500 people, completely independent of each other. If the analysis of the data will move at this pace, various talent shows may not lose their sense, but their performers will try to meet the basic criteria for the judging algorithm. The question is, would we be interested in watching something like that?
BBC in partnership with Nominet showed how to transform an SBC micro:bit into an IoT device. The first micro:bit was introduced last year and was designed for programming studies and STEM subjects. In March of 2016, micro:bit became available to British students and to the educational institutions and the plan was to supply up to a million of such devices. Micro:bit can be connected to such devices as the Arduino and Raspberry Pi and now the engineers figured out how to turn the micro:bit into a device to control gadgets equipped with IoT-sensors, offering the children a comfortable environment for their programming. This micro:bit will be used to solve utilitarian everyday tasks, for example, to analyze sensors to know when was the front door was opened at home or to check the air temperature in the classroom.
It was decided to automate online purchases in the UK using IFTTT task manager (“if this then that”). IFTTT system can create a logical chain of actions that can be associated with certain events. For example, if certain conditions are met, IFTTT will perform a certain action associated with those conditions. At Tesco, a British supermarket, it was decided to use this system to automatically buy certain goods. As an example, a customer can set up actions in his online personal account - if the price of some goods drops by 20%, then add 3 more items to one’s cart and notify via SMS or email. At the moment, Tesco offers the customers to take advantage of one of six sequences of actions established in IFTTT.
“Smart” pendant called “AutoDietary” paired with a mobile app with the same name, can count the number of consumed calories by the sound of chewing. There’s a microphone inside the necklace that is able to recognize the sound of eating, which, combined with the movements of the jaw and body vibration lets the pendant understand what kind of food you’re eating and how much.
The mobile application stores the information about nutrition into its database based on the data recorded and formulates recommendations on how to improve the diet. For example, if the wearer needs to consume more protein and vegetables. AutoDietary testing has show that the gadget works accurately in 84% of cases and more than 53% of testers said it is quite comfortable to wear.
UK scientists plan to create a database of MRI images for later study and diagnosis of diseases. As a part of this project, it is planned to make MRI scans of more than 100K people to use the data to predict the development of complex diseases (for example, cancer or Alzheimer’s) in early stages. A side goal of this project is to create the largest electronic database of human organs’ computer images.
At the annual Microsoft Build conference, Microsoft has presented its new project - an image recognition and description tool called CaptionBot. It allows users to upload photos to the site and it’s CaptionBot algorithm’s task to come up with a description, having recognized what is shown in the picture. The site uses Computer Vision API, Emotion API, and Bing Image API. Microsoft emphasizes that CaptionBot is just an experiment for now, which the company wants to run in an open mode to get the database for further improvements to the recognition algorithm.
Another interesting news related to image recognition is the emergence of Magic Pony technology that can recognize the image and improve the quality of the picture or video. Magic Pony algorithms are able to analyze a photo or video of poor quality and improve it, adding the lost fragments. For example, using the Magic Pony, you can restore damaged documents or photos.
Finally, a news from a funny category. The owner of a company that provides hosting services deleted all clients’ information from the servers, including the backup copies, with just one Linux command “rm -rf.” The Server Fault forum members already advised the negligent hosting provider to contact a company for hard drive data recovery. It is known that he was promised to get all data restored. (A side note: later it was discovered that it actually was a viral marketing campaign to promote the server management company.)