Learn More
The year ends where it began, with Microsoft winning several categories in the sixth annual ImageNet Large Scale Visual Recognition Challenge, beating out Google, Intel, Qualcomm, and others in many of the challenges, and more than halving the current record for correctly categorizing a set of 100,000 online images, with just a 3.5 percent error rate.
Microsoft dominates annual computer vision challenge.
Learn More
Getting the right computing power to run deep-machine-learning networks is one of the keys to advancing image recognition and artificial intelligence. By releasing its Big Sur AI hardware
designs, Facebook is making it easier for developers to build the resources they need to power deep machine learning. It’s also a nice complement to Google’s TensorFlow release, since that is just software.
Facebook open sources its AI
hardware, Big Sur.
Learn More
Researchers from MIT, NYU, and The University of Toronto unveil a new computer vision program based on Bayesian Program Learning (BPI). The BPI-based program beat humans in
correctly identifying handwritten characters by looking at just one example, a major feat, considering that a typical deep neural network learning method requires thousands and thousands of examples to get that accurate.
Computers beat humans in correctly identifying handwritten characters.
Learn More
Google releases its CloudVision API, which lets any developer add image recognition capabilities – everything from logo and face detection to the flagging of inappropriate content --
to their apps.
Google releases its image recognition technology to the world.
Learn More
Facebook demonstrates a new app based on its memory network technology that can answer questions via audio about what’s in a picture, enabling the visually-impaired to see what’s in their social media feed.
New Facebook app describes the content of pictures.
Learn More
Google releases its TensorFlow artificial intelligence (AI) engine to the world, enabling developers anywhere to use its technology for driving innovation around machine learning and its beneficiaries, including image recognition.
Google open sources its TensorFlow artificial intelligence technology.
Learn More
Pinterest unveils a new feature
that lets users zero in on any
product or object inside a picture
and immediately get search results listing similar pins, including where
to purchase.
Pinterest updates its image search.
Learn More
Tesla releases an over-the-air, downloadable software update that enables its Model S cars to partially drive themselves, thanks to a mix of image recognition-enabled cameras, radar, ultrasonic sensors, real-time data, and GPS. The ‘Autopilot’ service automatically keeps the car in a lane, changes lanes, watches out and automatically avoids collisions, and keeps a safe distance from other cars. It only works when the car is going more than 18mph and is optimized for highway driving. Autopilot also functions without a driver's hands on the wheel, though that is not recommended by the car company – yet. In the future, Autopilot will use image recognition, among other techniques, to help the Model S see stop signs and traffic lights.
Tesla’s ‘Autopilot’ enables its cars to partially drive themselves.
Learn More
Toyota and the Stanford Artificial Intelligence Lab (SAIL) announce the formation of a new research center, led by Fei-Fei Li, one of the world’s leading image recognition experts, to develop artificial intelligence solutions for self-driving cars.
Toyota and Stanford team up to make self-driving cars smarter.
Learn More
Baidu unveils a prototype of the portable Dulight device at its annual Baidu World conference in Beijing. The device is a small wearable camera that users clip onto their ear. Working in tandem with smartphones, the Dulight uses image recognition to analyze everything from food and currency to traffic signs and faces, then describes what it sees to the user. The Dulight is designed to be an assistant to the visually impaired.
Baidu’s wearable Dulight helps the visually impaired to see.
Learn More
of images online don’t have any identifying captions or hashtags
The majority of online and social media images don't have any identifying captions or hashtags, making it hard for companies to effectively monitor images relevant to their brands. By applying image recognition technology, visual listening tool Mantii helps brands find these heretofore lost images. Once found, the images are analyzed to deliver consumer insights to brands, as well as initiate social activation with customers, fans, and content creators.
80%
Mantii uses image recognition to alert brands when images relevant to them are posted to social media.
Learn More
Oluleye Hezekiah Babatunde, a computer science PhD candidate at Australia’s Edith Cowan University, develops a program that uses image recognition to determine the species of a plant just by looking at a picture of it. This can save days and weeks for farmers looking to monitor their crops, since previously the only way to identify a plant species was to send a leaf sample to a botany lab.
"Shazam for plants" helps monitor crops.
Learn More
Google releases ‘Deep Dream,’ a program that lets you upload any photo and let the search giant’s image recognition software analyze it and create new art based on what it sees. New sites such as Psychic VR Lab and DeepDreamGenerator crop up to make the process easy.
'Deep Dream' turns computers into surreal artists.
Learn More
Facebook demonstrates an algorithm that correctly identifies people’s faces, even when they’re partially covered, 83 percent of the time. It does this by relying on other identifying information in images, including body type, clothing, pose, and more.
Algorithm recognizes people, even when their faces are covered.
Joffrey Baratheon
THE LOSER
Learn More
Daenerys Targaryen
To demonstrate an algorithm that determines which pictures people are most likely to remember, a team of researchers at MIT’s Computer Science and Artificial Intelligence Lab fed their software images of characters from 'Game of Thrones.' The winner? Daenerys Targaryen. In 10th place, Joffrey Baratheon. Turns out the bad guys do finish last.
THE WINNER
MIT software picks the most memorable 'Game of Thrones' characters.
Learn More
Available as a smartphone app and on the web, the search giant’s new standalone photo manager uses image recognition to automatically file users' picture collections by faces and categories such as food, cars, skylines, beaches, and more.
Google Photos automatically categorizes your pictures.
Learn More
Upload any image, and the Wolfram Language Image Identification Project site at imageidentify.com will tell you what’s in the picture in just one word.
Wolfram’s new tool succinctly identifies any picture you submit to it.
Learn More
A research team at Carnegie-Mellon University demonstrates a long-range iris scanner that snaps pictures of a subject’s eyes and uses image recognition to determine the person’s identify. The demo focuses on a law enforcement scenario, since the system works up to 40 feet away.
Iris scanner identifies people up to 40 feet away.
Learn More
Vincent Van Gogh’s ‘Starry Night,” at the Museum of Modern Art in New York, remains the most popular.
Swedish startup Screenlab releases ArtNow, an app that uses image recognition to scour for shared pictures of artworks on Instagram, then categorizes them by title, artist, museum, style, popularity, and more.
Art Now for iOS tracks trending art on Instagram.
Download PDF
Humans in the experiment correctly identified the content of images 94.9 percent of the time
Microsoft’s deep-learning-based computer correctly identified the content of images 95.6 percent of the time.
“Nevertheless, we believe our results show the tremendous potential of machine algorithms to match human-level performance for many visual recognition tasks.”
“While our algorithm produces a superior result on this particular dataset, this does not indicate that machine vision outperforms human vision on object recognition in general,”
explained Microsoft's researcher.
Microsoft researchers in Beijing announce the development of a computer system that beats humans in image recognition accuracy by .5 percent. The milestone was accomplished on one particular set of images that have been used in tests before, however.
94.9%
95.6%
Computer beats humans in identifying the contents of photos.
Learn More
At the CES 2015 show in Las Vegas, the Japanese electronics company demonstrates kitchen appliances with cameras in them, including a refrigerator that uses image recognition to determine the expiration date of food, along with suggested recipes to cook that food before it spoils. Also: a camera-enabled mirror that automatically analyzes and keeps track of your skin’s condition, recommending relevant beauty products and regimens along the way.
Panasonic unveils smart home appliances that can see.
With big names such as Google, Facebook, and Microsoft opening their artificial intelligence technologies to developers, it’s clear that serious interest in the field of image recognition is gaining steam. Expect plenty of additional headlines, milestones, and more in the next 12 months.
2016 to bring increased innovation
in the field of image recognition.
‘Deep Dream’ turns computers into surrealist artists
Facebook open sources its AI hardware, Big Sur.
Microsoft dominates annual computer vision challenge.
Computers beat humans in correctly identifying handwritten characters.
Google releases its image recognition technology to the world.
New Facebook app describes the content of pictures.
Google open sources its TensorFlow artificial intelligence technology.
Pinterest updates its image search.
40ft
Tesla’s ‘Autopilot’ enables its cars to partially drive themselves.
Baidu’s wearable Dulight helps the visually impaired to see.
Toyota and Stanford team up to make self-driving cars smarter.
'Mantii uses image recognition to alert when images relevant to them are posted to social media.
“Shazam for plants” helps monitor crops.
Algorithm recognizes people, even when their faces are covered.
MIT software determines which 'Game of Thrones' characters are the most memorable.
Google Photos automatically categorizes your pictures.
Wolfram’s new tool identifies any picture you submit to it.
Iris scanner identifies people
up to 40 feet away.
Art Now for iOS tracks trending art on Instagram.
With records smashed, new services launched, and many big names stepping up their activities in the space, image recognition continued to grow in 2015.
2015
Image Recognition: Year In Review
Computer beats humans in identifying the contents of photos.
Panasonic unveils smart home appliances that can see.
December
November
October
September
August
July
June
May
April
March
February
January
Here are some of
the highlights.
Image Recognition