Monthly Archives: September 2018

Google Fonts launches Japanese support

Posted by the Google Fonts team

The Google Fonts catalog now includes Japanese web fonts. Since shipping Korean in February, we have been working to optimize the font slicing system and extend it to support Japanese. The optimization efforts proved fruitful—Korean users now transfer on average over 30% fewer bytes than our previous best solution. This type of on-going optimization is a major goal of Google Fonts.

Japanese presents many of the same core challenges as Korean:

  1. Very large character set
  2. Visually complex letterforms
  3. A complex writing system: Japanese uses several distinct scripts (explained well by Wikipedia)
  4. More character interactions: Line layout features (e.g. kerning, positioning, substitution) break when they involve characters that are split across different slices

The impact of the large character set made up of complex glyph contours is multiplicative, resulting in very large font files. Meanwhile, the complex writing system and character interactions forced us to refine our analysis process.

To begin supporting Japanese, we gathered character frequency data from millions of Japanese webpages and analyzed them to inform how to slice the fonts. Users download only the slices they need for a page, typically avoiding the majority of the font. Over time, as they visit more pages and cache more slices, their experience becomes ever faster. This approach is compatible with many scripts because it is based on observations of real-world usage.

Frequency of the popular Japanese and Korean characters on the web

As shown above, Korean and Japanese have a relatively small set of characters that are used extremely frequently, and a very long tail of rarely used characters. On any given page most of the characters will be from the high frequency part, often with a few rarer characters mixed in.

We tried fancier segmentation strategies, but the most performant method for Korean turned out to be simple:

  1. Put the 2,000 most popular characters in a slice
  2. Put the next 1,000 most popular characters in another slice
  3. Sort the remaining characters by Unicode codepoint number and divide them into 100 equally sized slices

A user of Google Fonts viewing a webpage will download only the slices needed for the characters on the page. This yielded great results, as clients downloaded 88% fewer bytes than a naive strategy of sending the whole font. While brainstorming how to make things even faster, we had a bit of a eureka moment, realizing that:

  1. The core features we rely on to efficiently deliver sliced fonts are unicode-range and woff2
  2. Browsers that support unicode-range and woff2 also support HTTP/2
  3. HTTP/2 enables the concurrent delivery of many small files

In combination, these features mean we no longer have to worry about queuing delays as we would have under HTTP/1.1, and therefore we can do much more fine-grained slicing.

Our analyses of the Japanese and Korean web shows most pages tend to use mostly common characters, plus a few rarer ones. To optimize for this, we tested a variety of finer-grained strategies on the common characters for both languages.

We concluded that the following is the best strategy for Korean, with clients downloading 38% fewer bytes than our previous best strategy:

  1. Take the 2,000 most popular Korean characters, sort by frequency, and put them into 20 equally sized slices
  2. Sort the remaining characters by Unicode codepoint number, and divide them into 100 equally sized slices

For Japanese, we found that segmenting the first 3,000 characters into 20 slices was best, resulting in clients downloading 80% fewer bytes than they would if we just sent the whole font. Having sufficiently reduced transfer sizes, we now feel confident in offering Japanese web fonts for the first time!

Now that both Japanese and Korean are live on Google Fonts, we have even more ideas for further optimization—and we will continue to ship updates to make things faster for our users. We are also looking forward to future collaborations with the W3C to develop new web standards and go beyond what is possible with today's technologies (learn more here).

PS - Google Fonts is hiring :)

Introducing the Kaggle “Quick, Draw!” Doodle Recognition Challenge

Posted by Thomas Deselaers, Senior Staff Software Engineer and Jake Walker, Product Manager, Machine Perception

Online handwriting recognition consists of recognizing structured patterns in freeform handwritten input. While Google products like Translate, Keep and Handwriting Input use this technology to recognize handwritten text, it works for any predefined pattern for which enough training data is available. The same technology that lets you digitize handwritten text can also be used to improve your drawing abilities and build virtual worlds, and represents an exciting research direction that explores the potential of handwriting as a human-computer interaction modality. For example the “Quick, Draw!” game generated a dataset of 50M drawings (out of more than 1B that were drawn) which itself inspired many different new projects.

In order to encourage further research in this exciting field, we have launched the Kaggle "Quick, Draw!" Doodle Recognition Challenge, which tasks participants to build a better machine learning classifier for the existing “Quick, Draw!” dataset. Importantly, since the training data comes from the game itself (where drawings can be incomplete or may not match the label), this challenge requires the development of a classifier that can effectively learn from noisy data and perform well on a manually-labeled test set from a different distribution.

The Dataset
In the original “Quick, Draw!” game, the player is prompted to draw an image of a certain category (dog, cow, car, etc). The player then has 20 seconds to complete the drawing - if the computer recognizes the drawing correctly within that time, the player earns a point. Each game consists of 6 randomly chosen categories.
Because of the game mechanics, the labels in the Quick, Draw! dataset fall into the following categories:
  • Correct: the user drew the prompted category and the computer only recognized it correctly after the user was done drawing.
  • Correct, but incomplete: the user drew the prompted category and the computer recognized it correctly before the user had finished. Incompleteness can vary from nearly ready to only a fraction of the category drawn. This is probably fairly common in images that are marked as recognized correctly.
  • Correct, but not recognized correctly: The player drew the correct category but the AI never recognized it. Some players react to this by adding more details. Others scribble out and try again.
  • Incorrect: some players have different concepts in mind when they see a word - e.g. in the category seesaw, we have observed a number of saw drawings.
In addition to the labels described above, each drawing is given as a sequence of strokes, where each stroke is a sequence of touch points. While these can easily be rendered as images, using the order and direction of strokes often helps making handwriting recognizers better.

Get Started
We’ve previously published a tutorial that uses this dataset, and now we're inviting the community to build on this or other approaches to achieve even higher accuracy. You can get started by visiting the challenge website and going through the existing kernels which allow you to analyze the data and visualize it. We’re looking forward to learning about the approaches that the community comes up with for the competition and how much you can improve on our original production model.

Acknowledgements
We'd like to thank everyone who worked with us on this, particularly Jonas Jongejan and Brenda Fogg from the Creative Lab team, Julia Elliott and Walter Reade from the Kaggle team, and the handwriting recognition team.

Source: Google AI Blog


High… Eight? The octopus slap heard ‘round the world and other searches this week

Riveting testimony on Capitol Hill, a new name for a favorite chain, and a snack that slaps back (and surprises an unsuspecting kayaker). Here are a few of the week’s most searched trends, with data from the Google News Lab.  

All eyes on the Senate

On Thursday, Supreme Court nominee Judge Brett Kavanaugh testified alongside Christine Blasey Ford in front of the Senate Judiciary Committee on Capitol Hill. As the testimony began yesterday morning, searches for the Senate Judiciary Committee spiked by 4,900 percent in the U.S., propelling search interest in the committee to its all-time high since 2004.

Hold the donuts

Dunkin’ Donuts dropped the “Donuts” and announced a rebrand to what most of us already call it: “Dunkin’.” Most related searches came from New England states Rhode Island, Massachusetts, Connecticut, New Hampshire, and Maine, which comes as no surprise since the company was founded in Quincy, MA. And while Weight Watchers also rebranded this week (to “WW”), search interest for “Dunkin donuts name” was 150 percent higher than “weight watchers name.” Munchkins, anyone?

Voting 101

Tuesday was National Voter Registration Day and above all, the U.S. wanted to know how. Trending questions included, “How to register to vote?” “How to register to vote online?” “How do I check my voter registration?” and “How do you know if you are registered to vote?” For all of you who helped “how to register” spike by more than 3,600 percent, Search can help. If you search “how to register to vote,” a tool at the top of the results page will help you find information about the voter registration requirements and processes in your state.

LMGTFY

Someone had a birthday this week... and we’re here to answer some of the top trending questions: “Is today Google’s birthday?” “How old is Google?” and “When was Google born?” We celebrated our big 2-0 on September 27 (although the exact date is up for debate), giving us the perfect excuse for a #ThrowbackThursday. Check out 20years.withgoogle.com, where you can dig into historical trends, like how we searched for candy in 1999 more than any other year, or how we have always searched for dogs more than any other animal (sorry, cat people).

Seal vs. Octopus vs. Kayaker

While testing a new GoPro camera, a kayaker in New Zealand was slapped in the face with an octopus by a seal (yes, you read that right). While the kayaker clearly took a hit, it’s been a close fight between the two sea dwellers in Search. ”Seal slap” started trending earlier than “octopus slap,” but the latter has been more popular over time, spiking 4,900 percent in the U.S. It turns out the seal was probably trying to pull off a tentacle as a snack, so maybe we call this one a tie.

P.S. Your humble High Five contributor was so, um, struckby this tale that she wrote a little something in haiku format. You're welcome:

Paddlin’ a kayak
Got slapped with an octopus
A tasty seal meal

Shawn Mendes collaborates with YouTube Music to release self-titled Artist Spotlight Story: ‘Shawn Mendes’

YouTube is thrilled to debut "Shawn Mendes," its latest Artist Spotlight Story, created in collaboration with international pop phenomenon Shawn Mendes. The Spotlight Story gives fans an authentic and intimate look into Shawn’s life following the launch of his latest self-titled album. Mendes has grown to be one of the most popular musicians today with a massive fanbase that stemmed from his early YouTube uploads covering other artists.

"It was just about doing it for fun and posting videos,” Shawn recalls of his earliest YouTube uploads in "Shawn Mendes," the latest installment of the Artist Spotlight Stories series from YouTube Music. “I wasn't dreaming about stadiums, or Grammys. I was dreaming of my friends being like, 'That's very cool. I love that. You sounded great.' I just wanted to be that guy who sang and played guitar on YouTube.”



Directed by YouTube Creator, filmmaker, and longtime friend Casey Neistat, "Shawn Mendes" looks back on Mendes’ journey from uploading covers to YouTube in his suburban-Toronto bedroom to performing on the world’s biggest stages as he gets ready to release his self-titled third album. Casey follows Shawn on the road and documents his rise to stardom through revealing interviews that showcase Shawn’s hard-working nature and youthful resilience.

With 15.7 million fans subscribed to his official channel, the three-time Juno winner is currently one of the biggest musicians on the planet. He's racked up 2.2 billion views across YouTube in the past 12 months alone, with “Treat You Better” and “Stitches” both passing the billion-view mark — making him the youngest artist to have more than one video break nine digits. More recently, his 2018 anthem “In My Blood” charted on YouTube in over 15 countries and landed the singer three MTV VMA nominations.

YouTube Music’s Artist Spotlight Stories are intimate and personal video portraits of artists who are pushing the boundaries of creativity and connecting with fans in unique and powerful ways on YouTube. Recent YouTube Spotlight Stories include Janelle Monáe: "A Revolution of Love," J Balvin: "Redefining Mainstream," G-Eazy: "These Things Happened," and Camila Cabello: "Made in Miami." The series is meant to shine a light on artist journeys and how they have leveraged YouTube to be successful, in hopes that their stories inspire others in the music community to do the same.

Source: YouTube Blog


Next Junction: Explore Indian Railways with Google Arts & Culture

https://2.bp.blogspot.com/-RB7SQNGLq2k/W63x1FCQyZI/AAAAAAAAAzc/RH1YgyrRQk43fEXIk-1QwAqfrjjUDWLdACLcBGAs/s640/image1.png Simon Rein Program Manager Google Arts & Culture
Over 151,000 kilometres of track, 7,000 stations, 1.3 million employees and 160 years of history. Indian Railways is one of the most celebrated railway networks in the world. A few months ago, we celebrated the 400th Indian train station connecting to the internet with Google Station, our public Wi-Fi program. Today, we’re bringing Indian Railways’ heritage and sights to the entire world. The most gorgeous architecture, iconic trains and charismatic personalities of Indian Railways can be found on Google Arts & Culture.


The first rail journey in India, a 14-coach train from Bombay to Thane in 1853 ushered in a new era of an India connected by track, rendering previously remote villages accessible. A century and a half later, just as trains once opened passages across the subcontinent, Google Arts & Culture’s new project  “The Railways — Lifeline of a Nation” is making these passages accessible for the world to experience.


Anyone can now explore India’s railways in unprecedented detail with over 100 exhibitions that bring together over 3,000 images, over 150 videos and more than 150 iconic locations captured across India. Zoom into ultra-high resolution images made with our Art Camera, like maps of the East Indian Railways that the British used to connect Calcutta with the North West Provinces; get a 360 degree look around the workshops of cardboard rail model enthusiasts; or take a behind the scenes peek at Darjeeling loco shed.

Discover the architecture of India’s most iconic train stations, including this Victorian Gothic stunner—the Chhatrapati Shivaji Maharaj Terminus in Mumbai

Watch how a railway inspired the making of a hit Bollywood movie

Go behind the scenes and watch how Indian Railways employees, like veteran technician Ganey Khawas keep the trains running


We invite everyone to take an online journey with us to see the breathtaking sights of India’s railways on the Google Arts & Culture’s online platform and the free Google Arts & Culture mobile app on iOS or Android.

By Simon Rein, Program Manager, Google Arts & Culture

Next Junction: Explore Indian Railways with Google Arts & Culture

Over 151,000 kilometres of track, 7,000 stations, 1.3 million employees and 160 years of history. Indian Railways is one of the most celebrated railway networks in the world. A few months ago, we celebrated the 400th Indian train station connecting to the internet with Google Station, our public Wi-Fi program. Today, we’re bringing Indian Railways’ heritage and sights to the entire world. The most gorgeous architecture, iconic trains and charismatic personalities of Indian Railways can now be found on Google Arts & Culture.


The first rail journey in India, a 14-coach train from Bombay to Thane in 1853 ushered in a new era of an India connected by track, rendering previously remote villages accessible. A century and a half later, just as trains once opened passages across the subcontinent, Google Arts & Culture’s new project  “The Railways — Lifeline of a Nation” is making these passages accessible for the world to experience.


Anyone can now explore India’s railways in unprecedented detail with over 100 exhibitions that bring together more than 3,000 images, 150 videosand 150 iconic locationsacross India. Zoom into ultra-high resolution images made with our Art Camera, like maps of the East Indian Railwaysthat the British used to connect Calcutta with the North West Provinces; get a 360 degree look around the workshops of cardboard rail model enthusiasts; or take a behind the scenes peek at Darjeeling loco shed.

  • chivpatri

    Discover the architecture of India’s most iconic train stations, including this Victorian Gothic stunner—the Chhatrapati Shivaji Maharaj Terminus in Mumbai.

  • bollywoodrailway

    Watch how a railway inspired the making of a hit Bollywood movie.

  • fairyqueen-indianrailways

    All aboard The Fairy Queen- the locomotive inaugurated by the East Indian Railways in 1855 and among the oldest functioning steam engines in the world.

  • ganey-indianrailway

    Go behind the scenes and find out how Indian Railways employees, like veteran technician Ganey Khawas keep the trains running.

We invite everyone to take an online journey with us to see the breathtaking sights of India’s railways on Google Arts & Culture’sonline platform and the free Google Arts & Culture mobile app on Android and iOS.

Introducing new APIs to improve augmented reality development with ARCore

Posted by Clayton Wilkinson, Developer Platforms Engineer

Today, we're releasing updates to ARCore, Google's platform for building augmented reality experiences, and to Sceneform, the 3D rendering library for building AR applications on Android. These updates include algorithm improvements that will let your apps consume less memory and CPU usage during longer sessions. They also include new functionality that give you more flexibility over content management.

Here's what we added:

Supporting runtime glTF loading in Sceneform

Sceneform will now include an API to enable apps to load gITF models at runtime. You'll no longer need to convert the gITF files to SFB format before rendering. This will be particularly useful for apps that have a large number of gITF models (like shopping experiences).

To take advantage of this new function -- and load models from the cloud or local storage at runtime -- use RenderableSource as the source when building a ModelRenderable. 

 private static final String GLTF_ASSET = "https://github.com/KhronosGroup/glTF-Sample-Models/raw/master/2.0/Duck/glTF/Duck.gltf";

  // When you build a Renderable, Sceneform loads its resources in the background while returning
    // a CompletableFuture. Call thenAccept(), handle(), or check isDone() before calling get().
    ModelRenderable.builder()
        .setSource(this, RenderableSource.builder().setSource(
                this,
                Uri.parse(GLTF_ASSET),
                RenderableSource.SourceType.GLTF2).build())
        .setRegistryId(GLTF_ASSET)
        .build()
        .thenAccept(renderable -> duckRenderable = renderable)
        .exceptionally(
            throwable -> {
              Toast toast =
                  Toast.makeText(this, "Unable to load renderable", Toast.LENGTH_LONG);
              toast.setGravity(Gravity.CENTER, 0, 0);
              toast.show();
              return null;
            });

Publishing the Sceneform UX Library's source code

Sceneform has a UX library of common elements like plane detection and object transformation. Instead of recreating these elements from scratch every time you build an app, you can save precious development time by taking them from the library. But what if you need to tailor these elements to your specific app needs? Today we're publishing the source code of the UX library so you can customize whichever elements you need.

An example of interactive object transformation, powered by an element in the Sceneform UX Library.

Adding point cloud IDs to ARCore

Several developers have told us that when it comes to point clouds, they'd like to be able to associate points between frames. Why? Because when a point is present in multiple frames, it is more likely to be part of a solid, stable structure rather than an object in motion.

To make this possible, we're adding an API to ARCore that will assign IDs to each individual dot in a point cloud.

These new point IDs have the following elements:

  • Each ID is unique. Therefore, when the same value shows up in more than one frame, you know that it's associated with the same point.
  • Points that go out of view are lost forever. Even if that physical region comes back into view, a point will be assigned a new ID.

New devices

Last but not least, we continue to add ARCore support to more devices so your AR experiences can reach more users across more surfaces. These include smartphones as well as -- for the first time -- a Chrome OS device, the Acer Chromebook Tab 10.

Where to find us

You can get the latest information about ARCore and Sceneform on https://developers.google.com/ar/develop

Ready to try out the samples or have issues, then visit our projects hosted on GitHub:

Twenty years of building for everyone

Editor’s note: This blog is cross-posted from The Keyword



Twenty years ago, Google started with an ambitious goal to organize the world’s information and make it universally accessible and useful. At the heart, we want to build technology that helps as many people as possible, regardless of who you are or where you are in the world. As we celebrate our birthday this month, we’re reflecting on some of the progress we’ve made toward that end. The work is never done, but here are 20 highlights from the past two decades:

Finding answers … and making connections

1. Billions of people have used Google Search to find answers to (literally) trillions of questions every year—from “how to help my community” to “how to find home” to all the many small questions in between. With job search, we’ve helped connect 100 million people to job opportunities in 92 countries.

2. Google Maps has helped people find their way with driving directions in 240+ countries and territories, spanning 40 million miles of road—that’s 83+ trips to the moon and back. And by connecting people to 150 million places around the world, Maps helps communities and businesses grow. Most moving are the times when Maps has helped people find a sense of place in the world—from Rio’s favelas to one’s own memories.


3. People can now talk to their Google Assistant in more than 20 languages, and in some cases it can even keep up if you’re bilingual. You can ask about everything from fantasy football advice to help finding a parking spot, and do everything from meditate to order a coffee. In the car, the Assistant has helped people reach their destination on tens of millions of commutes, and has sent tens of millions of messages, helping people stay in touch while keeping their eyes on the road.

4. Translate helps over half a billion people ask for help, make new friends, and say "thank you" across 100+ languages. More than 143 billion words are translated every day—that’s more than 161,000 times the number of the complete works of Shakespeare.


Saving valuable time (and space!)

5. More than 500 million people use Google Photos every month, backing up more than 1.2 billion photos and videos per day. Photos has also freed up over 410 petabytes worth of space—that’s like more than 25 million 16GB devices—plus peace of mind knowing you'll always have room to capture more memories.

6. With the typing time reduced by Autocomplete in Search, we estimate people worldwide collectively save over 200 years of typing time per day!

7. Gmail’s Smart Compose, a new machine learning-powered experience that helps you write email faster, saves people from typing over 1 billion characters a week (to put that in perspective, that’s the equivalent of nearly 4 million tweets).

Helping you stay safe online
8. One billion people visit the Google Account each year to access settings to safeguard their data and privacy. 

9. Safe Browsing protects more than 3 billion devices from malware and phishing schemes, helping you browse the web with confidence. 

10. And Gmail blocks nearly 10 million spam and malicious emails every minute, helping you keep your email and data safe.

Giving people tools and platforms to grow

11. Each year for the past five years, our search and advertising tools have helped provide more than $100 billion in economic activity to businesses, publishers and nonprofits across the United States. And we’re inspired by the stories of local and small businesses, from John’s Crazy Socks to American Hats, who are using the web to grow.

12. Google Play has helped developers grow app businesses and reach users in 190+ countries and across more than 2 billion Android devices. From an app that helps blind people see to a game that creates art, these creators are doing amazing things on our open platforms.

13. Since the start of 2017, we’ve trained more than 30 million people around the world in a range of digital skills, helping them start and grow businesses, learn to code, and find new careers.


Expanding access to learning opportunities 

14. More than 25 million students worldwide are using Chromebooks in schools to share ideas, create projects, go on virtual field trips, and learn from each other and their teachers. 

15. Art lovers and history buffs have marveled at artifacts from 1,500+ museums across 70 countries in Google Arts & Culture. From Abramovic to Zhengming, that’s thousands of artworks and 6 million photos, videos, manuscripts and other documents at your fingertips. And people have met more than 78 million selfie matches from 650+ institutions with Art Selfie. 

16. People can access local versions of YouTube in 91 countries around the world across 80 languages—covering 95 percent of all internet traffic. And every day, people watch learning-related content over a billion times on YouTube.


Making an impact on a global scale
17. To help people in times of need, we’ve activated SOS Alerts to provide better access to emergency information in more than 200 crisis situations, and people have viewed Public Alerts—for things like storm warnings or hurricane evacuations—more than 1.5 billion times.

18. Since 2005, we’ve donated more than $1.5 billion to organizations working to help refugees and disaster victims, fight for equal justice, provide teachers with classroom equipment, and teach people new skills. And over the past four years Googlers have logged over a million hours (that’s 114 years’ worth!) volunteering in the communities where we live and work. 

19. People have used Nest thermostats to save 25 billion kWh of energy—roughly enough to power Ireland for a year. 

20. Thousands of developers have used TensorFlow, our open source tool for deep learning, to make farming more efficient in Japan and the Netherlands, predict wildfires and prevent deforestation, track whale migration and identify birdsong—and even detect cancer.  

 
Google’s name is based on a number—a one with 100 zeroes after it. When we went public in 2004, the offering contained a math joke about the irrational number "e." Oh, and we call our campus headquarters the “Googleplex,” which, if it were spelled differently, would be a one followed by a googol of zeroes. You could say we’re numbers people. 

But these numbers are different. They represent something incredibly meaningful—the billions of people who have posed a question, sent an email, opened a new tab, dragged a pin on a map, asked a big question. Billions of people who have found answers, gotten things done or started on a new adventure. Billions of people whose lives have gotten, just possibly, a little bit better or brighter thanks to something that we built. 

Everything we’ve done for the past 20 years has been built with you in mind, and we’re incredibly grateful for the opportunity you’ve given us to be a part of your lives. And two decades in, we’re even more dedicated to building products and services that make a difference for you.

Posted by Emily Wood, Editor-in-Chief, The Keyword

Beta Channel Update for Chrome OS

The Beta channel has been updated to 70.0.3538.34 (Platform version: 11021.28.0) for most Chrome OS devices. This build contains a number of bug fixes, security updates and feature enhancements. A list of changes can be found here.

If you find new issues, please let us know by visiting our forum or filing a bug. Interested in switching channels? Find out how. You can submit feedback using ‘Report an issue...’ in the Chrome menu (3 vertical dots in the upper right corner of the browser).

Geo  Hsu
Google Chrome