Scale to launch a mapping product to address the changing needs of its autonomous driving customers

Solving the varied challenges that arise in autonomous driving is an incredibly complex task, but even attempting to get started means ensuring you have quality data that’s accurate and well-annotated. That’s where Scale comes in, having identified early on that the AV industry would require annotation of huge swaths of data, including specialized LiDAR imaging. Now, co-founder and CEO Alex Wang tells me at TC Sessions: Mobility 2021 (ExtraCrunch subscription required) that it’s moving into mapping with a new product that’s coming later this month.

“Our role has continued to evolve,” Wang said, regarding how it works with its transportation industry partners, which include Toyota among many others. “You know, as we work with our customers, and we solved one problem for them around data and annotational data labeling, you know, it turns out they they come to us with other problems that we can then help solve as well around data management, we launched a product called Nucleus for that. A lot of our customers are thinking a lot about mapping, and how to deploy with more robust maps. So we’re built a product, I’m going to announce that probably later this month, but we’re helping to address that problem with our customers.”

Despite my prodding, Wang wouldn’t provide any specifics, but he did go into more detail about the challenges of mapping, and what’s lacking in existing maps available to companies working on integrating those with AV systems that include other signals, like sensor fusion and vehicle-to-infrastructure components.

“I think a big question for the overall space has been that historically, the industry has relied very, very heavily on mapping — we relied very, very heavily on very highquality, high definition maps,” he said. “The tricky thing about the world is that sometimes these maps are wrong, and how do you deal with that? […] How do you deal with kind of this challenge of robustness, or updates. Even, if you think about it, Google Maps, which is the best mapping infrastructure in the world, by a huge margin, you know they don’t update quickly enough for [human] drivers.”

Wang said that the challenge isn’t all that different from the one that Scale has been actively solving for most of its existence, which is that of the data flywheel. With autonomous driving, it’s of utmost importance to be able to collect and annotate data quickly and accurately, which results in ever better collection and annotation of future data, and more reliability for the assumptions the system is making about its environment.

“Figuring out how to deal with the real-time nature of how the world changes, is one really big, one really big component,” he said. While we still have to wait to see what exactly Scale has planned, it seems safe to assume that it’s all about building confidence in maps and mapping accuracy as a key ingredient in whatever they launch.

#autonomous-driving, #event-recap, #mapping, #mobility-2021, #tc, #transportation

0

Deepfake tech takes on satellite maps

While the concept of “deepfakes,” or AI-generated synthetic imagery, has been decried primarily in connection with involuntary depictions of people, the technology is dangerous (and interesting) in other ways as well. For instance, researchers have shown that it can be used to manipulate satellite imagery to produce real-looking — but totally fake — overhead maps of cities.

The study, led by Bo Zhao from the University of Washington, is not intended to alarm anyone but rather to show the risks and opportunities involved in applying this rather infamous technology to cartography. In fact their approach has as much in common with “style transfer” techniques — redrawing images in an impressionistic, crayon and arbitrary other fashions — than with deepfakes as they are commonly understood.

The team trained a machine learning system on satellite images of three different cities: Seattle, nearby Tacoma and Beijing. Each has its own distinctive look, just as a painter or medium does. For instance, Seattle tends to have larger overhanging greenery and narrower streets, while Beijing is more monochrome and — in the images used for the study — the taller buildings cast long, dark shadows. The system learned to associate details of a street map (like Google or Apple’s) with those of the satellite view.

The resulting machine learning agent, when given a street map, returns a realistic-looking faux satellite image of what that area would look like if it were in any of those cities. In the following image, the map corresponds to the top right satellite image of Tacoma, while the lower versions show how it might look in Seattle and Beijing.

Four images show a street map and a real satellite image of Tacoma, and two simulated satellite images of the same streets in Seattle and Beijing.

Image Credits: Zhao et al.

A close inspection will show that the fake maps aren’t as sharp as the real one, and there are probably some logical inconsistencies like streets that go nowhere and the like. But at a glance the Seattle and Beijing images are perfectly plausible.

One only has to think for a few minutes to conceive of uses for fake maps like this, both legitimate and otherwise. The researchers suggest that the technique could be used to simulate imagery of places for which no satellite imagery is available — like one of these cities in the days before such things were possible, or for a planned expansion or zoning change. The system doesn’t have to imitate another place altogether — it could be trained on a more densely populated part of the same city, or one with wider streets.

It could conceivably even be used, as this rather more whimsical project was, to make realistic-looking modern maps from ancient hand-drawn ones.

Should technology like this be bent to less constructive purposes, the paper also looks at ways to detect such simulated imagery using careful examination of colors and features.

The work challenges the general assumption of the “absolute reliability of satellite images or other geospatial data,” said Zhao in a UW news article, and certainly as with other media that kind thinking has to go by the wayside as new threats appear. You can read the full paper at the journal Cartography and Geographic Information Science.

#aerospace, #artificial-intelligence, #deepfakes, #mapping, #maps, #satellite-imagery, #science, #space, #tc, #university-of-washington

0

Google Maps adds street-level details in select cities, more colorful imagery worldwide

Google Maps is getting a significant update that will bring more detail and granularity to its map, with changes that encompass both natural features and city-level details alike. For the former, Google says it’s leveraged computer vision techniques to analyze natural features from satellite imagery, then color-coded those features for easier visual reference. Meanwhile, select cities including New York, San Francisco and London, will gain more detailed street information, like the location of sidewalks, crosswalk and pedestrian islands, for example.

These additions will help people better navigate their cities on foot or via alternative modes of solo transportation, like bikes and scooters, which some have opted for amid the pandemic in greater numbers. The supported cities will also show the accurate shape and width of a road to scale to offer a better sense of how wide or narrow a street is, in relation to its surroundings.

Image Credits: Google (before: left, after: right)

While the added granularity won’t include more accessibility features, like curb cuts for example, Google says that having the crosswalks detailed on the map will help in that area. The company also notes that Google Maps today displays wheelchair accessible routes in transit and wheelchair attributes on business pages.

The updated city maps won’t show up immediately in the Google Maps app, we understand. Instead, Google says the new maps will roll out to NY, SF and London in the “coming months.” The vague time frame is due to the staged nature of the release — something that’s often necessary for larger apps. Google Maps reaches over a billion users worldwide, so changes can take time to scale.

The company notes that after the first three cities receive the update, it plans to roll out more detailed city maps to additional markets, including those outside the U.S.

Meanwhile, users both inside and outside big cities around the world will benefit from the changes to how natural features are presented in Google Maps.

Image Credits: Google

Google utilized a color-mapping technique to identify natural features from its satellite imagery, looking specifically at arid, icy, forested, and mountainous regions. These features were then assigned a range of colors on the HSV color model. For instance, a dense forest will now appear as a dark green while patchier shrubs may appear as a lighter green. You’ll be able to differentiate between beaches and greenery, see where deserts begin and end, see how much land is covered by ice caps, see where snowcapped mountain peaks appear, or view national park borders more easily, among other things.

These changes will reach all 220 countries and territories that Google Maps supports — over 100M square kilometers of land, from bigger metros to rural areas and small towns.

Image Credits: Google

The update comes at a time when Google’s lead as everyone’s default mapping app is being challenged on iOS and Mac. While Apple Maps started out rough, a 2018 redesign and subsequent updates have made it a more worthy rival. Apple even took on Google’s Street View with its higher-resolution 3D feature, Look Around, which particularly targets big city users. More recently, Apple introduced a clever trick that allows you to raise your phone and scan the skyline to refine your location. And Apple is battling Google Maps’ explore and discovery features through its expanded, curated guides built with the help of partners. These updates have pushed Google to race ahead with improvements of its own in order to maintain its lead in maps.

Google says the new features and updates will roll out across Android, iOS and desktop in the months ahead.

#apps, #google, #google-maps, #mapping, #maps

0

Mapbox and SoftBank form joint venture to provide mapping tech to Japanese developers

SoftBank Corp. and Mapbox, the mapping data company that competes with Google and Here, announced that they have established a joint venture called Mapbox Japan.

The JV will provide Mapbox’s mapping platform, including APIs and data services, to developers in Japan. Between June 1 and September 30, Mapbox Japan will also provide up to three months of free support for organizations building COVID-19 related mapping services, including infection cases and statistical data, for developers in the country, which has relied on tracking virus clusters to limit the spread of infections.

Mapbox collects data from sources including government and commercial databases, and uses them in customizable AI-based APIs, SDKs and other products. Its clients have included Facebook, Snap, the New York Times, the Federal Communications Commission and automotive companies like Land Rover and Rimac.

Founded in 2010 by Eric Gunderson, Mapbox says its tech now reaches more than 600 million monthly users. SoftBank Vision Fund led Mapbox’s $164 million Series C in 2017. At the time, Gunderson told TechCrunch that part of the funding would be used to expand in Asia through SoftBank’s presence in regions including Southeast Asia and China.

Mapbox has operated in Japan since July 2019, though that was through partnerships with Yahoo! Japan and Zenrin, one of the country’s biggest mapping software companies. Zenrin also has a partnership with Google Maps, but early last year Google began reducing the amount of mapping data it uses from Zenrin, possibly to focus on building its own trove of mapping data in Japan.

Working closely with Zenrin opens potential new opportunities for Mapbox in Japan. Last year, Gunderson told Nikkei Asian Review that “we are going to be the number one mapping provider in all of Japan and we’ll be able to do this because we have the best data in all of Japan through our partnership with Zenrin.” The company plans to develop products for the Japanese market that include mapping services for industrial automation.

In SoftBank’s announcement, Eric Gan, SoftBank Corp. head of business development, said, “I am very excited to bring Mapbox’s technology to Japan to help enterprises enhance their existing mapping services while also creating new customizable location-based services and management tools. We are seeing a significant rise in demand for Mapbox’s products from retail, ride-share, hotel, office-sharing, payment, mobility and manufacturing industries.”

#japan, #mapbox, #mapping, #mapping-technology, #softbank-corp, #startups, #tc

0