Street-level Imagery as Open Data
This is a modal window.
The media could not be loaded, either because the server or network failed or because the format is not supported.
Formal Metadata
Title |
| |
Title of Series | ||
Number of Parts | 295 | |
Author | ||
Contributors | ||
License | CC Attribution 3.0 Germany: You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor. | |
Identifiers | 10.5446/43462 (DOI) | |
Publisher | ||
Release Date | ||
Language |
Content Metadata
Subject Area | ||
Genre | ||
Abstract |
| |
Keywords |
00:00
Open setLevel (video gaming)Slide ruleLecture/Conference
00:26
Open setDreizehnTable (information)PolygonOrder (biology)Level (video gaming)Point (geometry)Military baseComputer animation
00:50
Pauli exclusion principleDigital photographyWordUniform resource locatorSoftwareMereologyGeometryLecture/Conference
01:16
InformationPixelTimestampMultiplication signDigital photographyDigital photographyTwitterLevel (video gaming)GeometryDigitizingGroup actionHypermediaDifferent (Kate Ryan album)InformationForm (programming)Medical imagingWater vaporDerivation (linguistics)
02:04
Menu (computing)MaizeVideo gameSource codeInformationPixelTimestampDigital photographyOpen setWeb portalInformationFerry CorstenShared memoryHypermediaProjective planeTimestampRepository (publishing)Digital photographyService (economics)Different (Kate Ryan album)Open setElectronic GovernmentCartesian coordinate systemWeb portalCovering space
03:29
Open setWeb portalDigital photographyVariety (linguistics)Different (Kate Ryan album)File formatMedical imagingOpen setComputer fileStructural loadCategory of beingLecture/Conference
03:47
Open setWeb portalDigital photographyGraph (mathematics)GeometryResultantFile formatDifferent (Kate Ryan album)Shape (magazine)MereologyCategory of beingSoftware repositoryOpen setDigital photographyAxiom of choice
04:32
Digital photographyOpen setWeb portalSatelliteMultiplication signService (economics)Goodness of fitSource code
04:50
Perspective (visual)Digital photography1 (number)Lecture/Conference
05:11
Price indexOpen setDistribution (mathematics)Library catalogWeb portalDigital photographyOpen setDigital photographyForm (programming)DigitizingMedical imagingGeometryDescriptive statisticsComputer animation
05:54
Open setComputer-generated imagerySquare numberTexture mappingDigital photographyOpen setTesselationLevel (video gaming)Group actionGeometryInformationComputer animation
06:38
Digital photographyMedical imagingDigitizingDifferent (Kate Ryan album)Multiplication signData storage deviceUniform resource locatorRight angleAreaBack-face cullingWater vaporOpen setWebsiteDigital watermarkingLibrary (computing)
07:31
Digital photographyMathematicsPoint (geometry)Type theoryMatching (graph theory)Computing platformGraphical user interfaceLecture/ConferenceMeeting/Interview
08:05
Open setRepository (publishing)Form (programming)MappingComplete metric spaceLecture/ConferenceComputer animation
08:44
Digital photographyPoint (geometry)JoystickPoint (geometry)Medical imagingLevel (video gaming)TwitterSet (mathematics)Physical systemMathematical analysisDigital photographyTransportation theory (mathematics)Hard disk driveComputer animation
09:31
Level (video gaming)Hard disk driveMedical imagingUtility softwareSpur <Mathematik>Condition numberLecture/Conference
10:03
Digital photographyPoint (geometry)TriangleOpen setSign (mathematics)Utility softwareCondition numberSpur <Mathematik>TriangleMedical imagingOpen setWeb applicationLevel (video gaming)Bridging (networking)Internet service providerProjective planeCycle (graph theory)Validity (statistics)BitSparse matrixHard disk driveComputer animation
11:33
TriangleState of matterLevel (video gaming)Open setMedical imagingTransportation theory (mathematics)AreaData conversionSpur <Mathematik>Streaming mediaMobile appElectronic mailing listLecture/Conference
13:00
Motion captureArchaeological field surveySystem programmingLinear mapPerspective (visual)Medical imagingRight angleSystem administratorSoftware maintenanceVolume (thermodynamics)Degree (graph theory)BitSpur <Mathematik>Internet service providerLevel (video gaming)Transportation theory (mathematics)Open setPerspective (visual)MathematicsMatching (graph theory)TwitterLecture/Conference
14:24
Condition numberNumberMedical imagingAngleSelf-organizationOffice suiteEnterprise architectureCartesian coordinate systemLecture/ConferenceMeeting/Interview
14:49
Motion captureArchaeological field surveySystem programmingLinear mapPerspective (visual)Cartesian coordinate systemSelf-organizationEnterprise architectureInternet service providerOffice suiteMedical imagingAddress spaceLine (geometry)Motion captureDisk read-and-write headLevel (video gaming)PolygonSource codeComputer animation
15:29
Line (geometry)Level (video gaming)Disk read-and-write headPerspective (visual)PolygonReal numberLecture/ConferenceMeeting/Interview
16:03
Motion captureArchaeological field surveyLinear mapSystem programmingPerspective (visual)Mobile WebMedical imagingUniform resource locatorLevel (video gaming)Boundary value problemDifferent (Kate Ryan album)BuildingMobile appWeb 2.0Text editorOffice suitePlug-in (computing)AreaWebsiteRange (statistics)Open setMatching (graph theory)Projective planeOpen sourceComputer animation
17:16
Data typeSoftware developerOpen setMobile WebScripting languageRight angleFile viewerSoftware developerDemoscenePixelCartesian coordinate systemData structureMedical imagingOpen sourceProcess (computing)Level (video gaming)CodeLibrary (computing)Volume (thermodynamics)Open setLine (geometry)Lecture/ConferenceMeeting/Interview
18:37
Motion capturePairwise comparisonType theoryMobile WebLevel (video gaming)Multiplication signNumberCartesian coordinate systemDegree (graph theory)Medical imagingOpen setPairwise comparison
19:24
Dependent and independent variablesMappingType theoryTrailVulnerability (computing)Wave packetLevel (video gaming)Amsterdam Ordnance DatumLecture/Conference
19:53
Mobile WebPairwise comparisonType theoryMotion captureSource codeGoogle Street ViewLevel (video gaming)Google Street ViewSource codeMappingInternet service providerGeometryMultiplication signSelf-organization19 (number)MultiplicationLine (geometry)
20:39
Source codeGoogle Street ViewInternet service providerSelf-organizationMappingProduct (business)
21:02
Type theoryRoundness (object)Medical imagingPlanningLevel (video gaming)Mathematical analysisLecture/ConferenceMeeting/Interview
21:51
Source codeGoogle Street ViewBuildingType theoryMedical imagingContent (media)Category of beingLecture/ConferenceMeeting/Interview
22:11
Source codeGoogle Street ViewInformationElectric power transmissionGroup actionMedical imaging
22:42
Lecture/Conference
22:59
Texture mappingSource codeGoogle Street View
Transcript: English(auto-generated)
00:07
My name is Chris and I'm going to talk to you about Street level imagery as open data if I could change slides. I
00:21
Got stuck on my slide. Hang on. Okay, so When we talk about data There are many different kinds not all data is geospatial We know this very well. Sometimes we're opening a table based piece of data
00:43
Something that's a CSV and you need to do table joins in order to Join it to the map sure to hit it with points or polygons But most of us are here because we're talking about things like location data We're talking about spatial data geodata. We have all these different words for it and
01:02
most of all we like to talk about open data and This is an important part of the phosphor G world because you have the software you have the people But it's all built around making sense of this data, but we don't always think of photos as data
01:21
going back to old school photos something taking with the camera on film a Lot of these are now a form of data that's still accessible to a lot of us And when we use a digital camera or something on our telephone on the mobile phone With action cameras like a GoPro
01:41
You're getting a lot more than just the visual information so We can interpret the photographs to create information and we can do something like derive data from them Just by looking at them but one of the most important things for us in the geospatial world is
02:01
Looking at the geotag of an image. So people will do this with things like Twitter You'll see it with all kinds of different social media apps. You'll see it with things like flicker Where suddenly the photo becomes relevant to the map? But you also have more information inside of Something like a JPEG so you can get pixel values. You can get the timestamp which is very correlated to the geotag
02:27
and overall we call this the exit and exit was short for I forgot it now to But Essentially what this is is it's something that if you're working with photographs a lot. It's what you're actually analyzing and
02:44
if you're working with geotag photos then accessing the exif is Kind of the first step to actually opening up that photo and getting under the hood And when we want to take that data and share it we want to make it open
03:02
We're often dealing with open data portals. So this is very common for us when we're working on a project we say, okay I need data about Landcover I need data about Something in biogeography Where do I get it? So you're going off into a government website or somewhere that has a repository of data
03:23
so it's often a government service and When you access a government open data portal, you're often seeing a A variety of different file types available to you there Let's see I
03:41
See if I can get this to load my next images So here for example, we're looking at the Swiss open data portal and you can see categories. You can see keywords If you scroll down later, you also see things like different file types and they're not all shape files
04:01
Some of these are zips you have csv's JSON but it's not a geo JSON So it's not always geodata but geodata is definitely a growing part of these repositories and for many of us It's the reason we visit them You also have things like imagery inside of these these open data repositories
04:21
But often when we're thinking about imagery people are thinking about something aerial. So here's from the city of Buenos Aires you have aerial photography is one of the first results when I search for photography photography and That's also from a camera, but it's
04:41
It's something that's been around for a long time Like we expect that often is a government service that maybe your city or your country is Capturing annual aerial imagery or even someone like NASA is sharing satellite imagery and it's a public good but aside from this you have
05:01
Just regular photos just the ones from our perspective on the ground and it's interesting to look at how these can be shared as open data and how it How it gets interpreted so, let's see the example you see here
05:20
This is a photograph and it was taken by a Swiss photographer. You can see in the description He started I think this was the late 1800s was when he was capturing these So looks like his name was Adolf Braun beautiful photographs of Switzerland and These were digitized and they're now shared on the Swiss open data portal
05:43
These aren't necessarily geotagged. However, so you have a form of photography as open data But you don't have it as geo data. So there's a next step there and some people have taken these steps, but Let's bounce back to imagery as geo data
06:01
Another great example is open aerial map This imagery you see here is from Zanzibar. So there are groups in Zanzibar who are capturing drone imagery They share it here and people were able to tile it on to open street map For example, and then start mapping things like building outlines roads paths and other geographic features
06:23
so it's both imagery that already is geospatially associated as Well as a resource that you can use to actually drill down with Human eyes and start extracting more geospatial information from it. So we looked at the the photograph from Switzerland and
06:44
That's historical imagery. It's open data on the left here. You'll see from the National Library of Denmark another historical photograph That's being shared openly on their website and it's been digitized from a time before you could upload and geotag photos
07:01
but suddenly the difference here is Not only can you compare it with a modern image, which you see on the right But both of these are geotagged So both of these you can actually find on Mapillary there's a small watermark in the corner of each one of them and what's interesting is
07:21
one of the Mapillary ambassadors based in Copenhagen or the Copenhagen area he was able to take all the digitized photos and start trying to match them with the correct locations on the ground and The next step is you could then compare them with photos that another user Like Peter Neubauer of Mapillary was able to go out and capture with his mobile phone and
07:44
The matching starts to happen automatically at a point Which we'll get into More of a technological discussion here, but overall these are available on the same platform Side by side and treated as the same type of data
08:00
So they're totally from different eras. They're totally from different devices The first one on the left was never intended to be shared this way you were never supposed to pull it up in Chrome back when it was captured in something like 1890 But here it is so This concept of imagery being open data, especially imagery coming from government repositories or government resources
08:27
captured by government a Lot of it came to become a form of open data that we use to fix maps By a complete accident and If you know Bob Ross, he likes to call them happy little accidents and our happy little accident
08:45
Happened a few years ago so Steven works for the Vermont transportation agency and Steven Advertised on Twitter that he was just having fun Doing some analysis with data he had found
09:01
You can see it on the map and he said these points are from 2012 and 2013 Each photo is about 400 kilobytes But but uh, but the data set is almost 2 million points each one representing an image on Vermont's roads, so this is the highway system and he asked is this interesting to anyone?
09:24
so a shot in the dark a tweet in the open and I happened to see this and I said hey, I have an idea we'd love to Talk to you more about what you're doing with this imagery so he had a drawer full of hard drives and It does go back to 1998 which is a huge amount of data, but we took a couple years of this data
09:45
we were able to have it shipped to us at Mapillary and Because each one of these images was already geotagged We're able to start putting them on the map and stitching them together into a street level imagery experience And this is kind of totally different than they ever expected to do
10:03
originally, they were Just capturing these in utility trucks from the government and they'd use it to review the condition of the roadways But they didn't really have a larger purpose in mind as far as sharing it or letting other people repurpose it So this kind of led to a bit of a triangle the open data triangle that I work in a lot with
10:27
my own experiences so Mapillary is not a provider of street-level imagery we are more of a bridge between people who go out with cameras and capture it and
10:41
People who want to consume it for some special purpose or project sometimes it's the same people but they need a bridge between their camera or their hard drive and their actual web applications or desktop GIS Sometimes it's the open street map community and they're repurposing images for example from Vermont
11:01
so on open street map you have access to this imagery as well as data derived from it and Ideally we connect that triangle and one final step where governments actually start to embrace OSM as a Valid base map rather than something that's much older or more sparse or less frequently updated
11:23
So hopefully they're giving out imagery. We're making it available people are using it creating a map and completes It continues the cycle so that it's worthwhile to keep capturing more imagery keep updating that map and keep using the map
11:41
so Vermont's the first example of What started this conversation about governments using Mapillary to share imagery as open data? But we want to ask. Okay, who else is doing it? Well very quickly. There were many more and To this day that list is increasing So in North America, you have several different states
12:00
These are all state departments of transportation As well as you have cities, for example the city of Detroit, which is also editing open street map from the Mapillary imagery and then reusing it so it's that triangle we talked about and In Canada as well. You have cities like Saskatoon or st. John's and it goes on through Europe
12:26
we have several in Oceana and Australia and Even in South America the National Transport Agency of Brazil which drives Vast amounts of roads through a very large country They also are using cameras to capture this roadway. So suddenly this is becoming open data that available as well
12:46
so imagine editing open stream app in the Amazon in places that are very hard to reach but there are some roads that go through there and suddenly you have Not just one but many many hundreds of images
13:00
Just in one small stretch of a road on the ground and frequently up-to-date So you can see some example images on the right of what it looks like on the ground in each of these places these are each taken from Mapillary and They're all from the user account of the administration. You'll see on the left side there
13:24
so we talked a little bit about v-trends and In Vermont they had a purpose for capturing that imagery originally which was roadway maintenance And a lot of people ask, you know when this huge volume of imagery does come into Mapillary suddenly in their country Maybe you're living in Lithuania and one day you log in to Mapillary or open street map
13:45
and you find that there's imagery available in 360 degrees of all the roads and It's only from last year Which is very different from many other street-level providers that are five years old or three years old And you're asking why are they doing this?
14:01
What's the incentive for them to actually go out there with a camera and what's the incentive to upload it? So from a perspective of government transportation agency The imagery is helping to document Sometimes the work that you're doing Maybe you're putting in fiber-optic cables along a roadside and you need to have imagery that shows that you finish covering them back up with
14:25
pavement And then when you also are doing something like analyzing the road condition across the country You're in a GIS tool and you need to quickly to take a look at What mile marker or kilometer marker number 75 looks like on the ground?
14:41
You can get an image that you know is located near that marker and has a camera angle facing toward it So this is hard to do in practice many of these organizations had Applications that were just kind of strung together by their own office but less than satisfactory or we're using enterprise applications that were very expensive but not really tailored to their needs and
15:09
A lot of them were also looking at commercial street imagery providers and they'd say well, it's outdated We want to just see what this house looks like at this address, but the image is from 2012 We're going to go out and capture our own imagery
15:23
so among these reasons and many more All these agencies are looking for a way to just enrich their GIS by actually getting a visual on the ground Like we often think of the map Similar to the quote the map is not the territory we think of it in our heads you sometimes
15:42
Find yourself dreaming of it overnight the polygons and the the road lines running through and you realize Okay, that's me thinking about the city. I live in but from this Theoretical perspective that's me and QGIS instead of looking at it the way it is when you walk down the street
16:01
so Sometimes when you're editing the map you actually want that real-world perspective and this is it There are other uses as well things like automated data collection or even manual you can step through in your office and mark The location of crosswalks just by seeing images that are geotag near them
16:22
so those uses They cross a lot of different boundaries and there's always new purposes being invented But one of the most important pieces is also inventing new tools So you can see here the QGIS plug-in this is called go to Mapillary
16:40
so this was made by an architect in Italy Enrico Farraguti and he was making this as a personal project because he needed to see what the buildings looked like in an area he was working and Suddenly it was shared as an open source plug-in and now you can pull Mapillary into this GIS tool and add your own imagery as well
17:00
You can view it on the Mapillary website you can view it across a wide range of open street map editors As well as any kind of proprietary GIS tools like ArcGIS you can view it in the here map creator and many other custom web and mobile applications so the goal here overall is
17:23
Getting the imagery online is one thing getting people who have huge volumes of imagery online is another and then making it available to reintegrate into creative scenarios That's where you get developer resources So in my own work, I'm often helping people who have imagery or know where the imagery is that they want to use
17:45
many of these are government entities and they're saying okay, we have it, but we want to use in a very specific way and So using resources like this You're able to then integrate it. So at the top right you see Open structure for motion open SFM
18:03
So this is an open source library that Mapillary provides and it helps you recreate 3d scenes based on the pixels in that street level imagery and The bottom right you see just the very basic example code just for using the Mapillary viewer so you can throw this inside of a JavaScript application and
18:23
With these few lines you already are able to visualize an image and connect it to your map So the goal is to make this very easy Alongside with how easy we like to make it for uploading that imagery. So finally, I want to talk about the future
18:45
of Mapillary and as the street level imagery as open data So one is we want to enable more ways to capture So there's a lot of imagery out there That's not getting uploaded because it's not geotagged or people just don't know how to connect it to the methods for uploading
19:03
But this includes just expanding the number of mobile phone applications that are able to submit images to Mapillary connecting automotive sensors and Also, just enabling better support for more advanced cameras 360-degree cameras. For example the more historical imagery we get we also can get more comparisons over time, which
19:26
can lead to a lot of new learnings and We also will get imagery on new road types not always just highways or city streets But even in residential areas train tracks footpaths
19:40
so and that leads into really a question of mapping what's Beyond commercially viable and governments often take responsibility for this mapping vulnerable communities mapping in developing countries Or even grassroots level mapping led by citizens and communities rather than government industry such as the map 2020 initiative
20:05
so overall, I just want to encourage you to take a look at this timeline and Ask about the future of who the providers of street level imagery are In the past it's been proprietary you can see Like Google Street View in 2007, but people were doing this even before then like 1998 in Vermont
20:25
Just without the idea of what it could become so now as we go into 2019 we see there are many many many sources including 2018 across multiple cities Organizations like bike Ottawa who does bike advocacy
20:44
GeoChica's has many mapping products going on. All of these are new providers that are independent of any kind of large company And in 2020, we want to look at who else can contribute who else can consume the imagery and how it will be shared
21:00
Thank you Thanks, Chris. So we have about maybe three minutes for questions So if you did plan to go to another session We'll have to kind of juggle the two so stick your hand up and I'll bring the mic around if you have a question
21:22
for Chris Questions for Chris
21:44
Did you do any regional analysis with the images, you know like on a map to see like, okay, there's this type of Buildings or this type of roads have used it in this way like with image classification
22:04
Yet we did we've done some really interesting things about the content of the images Roughly you can actually just tell what percentage of an image is classified in a certain category like buildings roads or road markings So one exciting example that I want to share more work on is just mapping urban footprints in the Galapagos Islands
22:26
You can start to classify if the imagery is Containing more urban things like wire groups power lines or roads or if it's more like vegetation and in bare earth So there's a lot of that data available But we also don't specifically provide that as any kind of information. We just make it available for someone to work with
22:56
Thank you. Take care