"OpenStreetView." By: Alex Ilisei and Martijn van Exel. >> So we're almost at the end of this conference. I've had a pretty amazing time. How about you all? [Applause] Yeah, I've been to a lot of OSM events, probably more than a dozen conferences and I guess -- who of you has been to Kevin's talk earlier? Kevin. Did you notice how many people were new to the conference? And one of the reasons there was almost half of the people were first time at the conference, and I think why that's important is OSM is growing really fast but not everybody actually starts mapping, so I think we need more ways to get people to go from signing up to actually starting mapping. And I think all the people presenting today can help with that also. So here's Alex, my colleague Alex, we're going to be doing this talk together switching back and forth. We have quite a few slides, so let's get started. So who of you has used street level, some sort of street level imagery for OSM before? I would say that's about half of you. So that's what OpenStreetView does and I'll go real quickly into what that is. So you have an example of OpenStreetView here, it has all the information of it, this one I took myself from my car using OpenStreetView. One of the things you can do is map bison crossing. [Laughter] This happens when you live in Utah. I don't know if there's bison here, but they're everywhere there. So you have to be very careful. That's not the main purpose, though. I mean bison are nice, but we all know how to map from aerial imagery. Missing a lot of perspective; right? So you can't see -- these are the same locations from the top down and from the front view and given none of those things from the top. So you need really both to do proper mapping. Another reason I think is, like, there's more eyes, we get more eyes on the world. And this is where the new map comes in; right? Lots of folks will drive around or bike around and take pictures. But only very few will go and map. I think there's many more people you can get to drive than you can get to map. Even though however much you want to change that. So I think street level has the potential to grow OSM at a pretty fantastic rate because it allows us to have much more information when you're not at the scene. So enter OpenStreetView; right? What is it? It's a free and open platform, and I'll go into all of these things a little bit later and Alex will as well. Platform that is created 100% for OpenStreetMap. There's no other purpose at all. This is it. It's built for you and built by you because we're open sourcing everything. And then the next thing is going to be -- Alex is going to be talk about what are the components of the platform and what are the things that you can be using today. So Alex? >> Hi, is this on? Can you hear me? >> Yeah. >> So this is high level view of OpenStreetView. So we start with the open source apps for iOS and Android that do the actual recording. And then the -- we build some tools for people that don't use different cameras like go on pros or all kinds of cameras that record photos. We have the back end that basically stores and also does the image processing of everything. And we feed the image and the results of the processing to basically OpenStreetView.org where you can see all the recordings to a dedicated editor tool. This is a desktop tool we'll get into this in a little bit. A JOSM plug in and of course this ends up in OSM. So I'll go through some of the components. This is the mobile app basically this is what you see when you open it. And all the purple stuff is basically coverage in your area. And you can tap on the roads to see the pictures that are taken on that location. So you can see here if you tap on a road, you can see different tracks from people that went there. And you can also see an individual track. So here's an example of the so it's a cool time lapse if you see after the recording; right? Cool so this is how the recording stream looks like. You can have a big camera view or big map view of the recording. And of course after you finish, you have to upload the data. You're going to need Wi-Fi for that, so we're not live -- we're not streaming anything. You can also record in offline mode. But when you upload, of you to have Internet connection and in the right there's the profile page with ranking and all the good stuff. Yeah, one thing that I'm really proud of is the pied piper. I'm sure you get the reference. When we started we get the JPEG files and we did a little magic and get less than half of JPEG files. It's open source. Go check it out. It's really cool. Yeah, upload tools for camera users, it looks like this is the desktop app, so you can upload all kinds of data. This is the website. OpenStreetView.org. This is also the coverage, leaderboards, and all the good stuff. One other thing that I'm proud of is the map matching. So here I think you're familiar with this look. All the GPS traces; right? Made them look pretty, ended up with this. So now if you tap on a location, you can see the green dot there. You'll see all the tracks of people that went there. So you see different pictures at the approximate same location, and you can scroll through. This is the detail page, so you can see the actual tracks that navigate through them. Of course the leaderboard, it's crazy what people do just to get in front. There's someone here in the audience that took a helicopter ride just to -- [Laughter] This is the profile page where you can see all the statistics and your tracks and all of that. We do use OpenStreetMap account exclusively, so you'll need to have an OpenStreetMap account if you want to use OpenStreetView. I guess that's not an issue. And one important aspect that you remain in charge of your recordings of your tracks; right? So you can always download all of your own data, do what you want with them and also you can delete your account and wipe everything if you don't want to show us anymore. >> Thanks, Alex. So we build a few mappers specific things. So these are things that are all under heavy development and probably will remain under heavy development just because we want to gather a lot of feedback. So this JOSM plugin is very basic right now. So what it does is take all the locations where we only have photos, and it will just plot them with little camera icons on your JOSM map view, which is cool, but it also gets -- once the coverage gets up there, it gets very clever. And we're working on ways to make that all, like, clean -- cleaner looking and better usable. But you can download this now and work with this. Of course make sure that there is some coverage in your area. So download the app first and create that. JOSM plugin. This is kind of -- it's in the JOSM plugin storm. I don't know if you can call it a store. Probably not. It is a plugin repository, search for OpenStreetView, and you'll get stuff like this. This picture is something I took a while ago. This is from my GoPro where I was on my bike, you can see my hands there on the Handlebars. So that's also a possibility. You don't have to be in a car, even though we like the driving use case a lot. So this is what you see. We would love to get your thoughts on this and see how we can develop a little bit more into something truly awesome. You can filter on date and also adjust your own stuff, which I think is important. Talking about seeing your own stuff, we do the OpenStreetView editor, which is that's kind of a new component, I want to spend a little bit of time on that. This is a desktop tool that you have to download. So it's something new that you'll need to kind of learn how to work with. It's very easy to use, though. So you can -- it's built with QT, so it's cross platform easy and also open source. And we think it is going to be efficient to map stuff based on your drives. So that's what it is right now. This is a bit what it looks like, and you get a view of your own traces and where they are. With a little bit of on those whether you get into OBD2 with how many things detected on it and with some other metadata to help you select the best one. So then you get to the video track and you're going to scrub through that a little bit using the -- I'll get my laser pointer. So there's -- no, I don't get my laser pointer. So you can scrub through those, and you see at the bottom the line there represents all the stuff that was detected in terms of right now we detect U.S. speed limit signs that will be -- that will expand in the very, very near future. And if you click on any of those, and you see the little part of the image that wasn't detected as a speed limit and can really quick on those and the really interesting stuff that happens actually on the right, it will download -- because we map match everything already, so we know which -- what the OSM that belongs to this. And then we can pull the speed limit from there. We can compare it. And if it doesn't match, then we can -- you can apply that actually use the JOSM remote control. So you can apply the speed limit directly to -- directly to the OSM data and then upload it using JOSM. So the tool itself doesn't upload data, it uses the JOSM remote control to communicate, and then you can upload the changes after that. So that's kind of how that works. Speed limit, you see there that in this case it wasn't in OSM yet, so you can and OpenStreetView has it, so you can see the warning icon. It can apply the change, and then it will go to JOSM, so you can fact check it first. And then the data will be in OSM. So pretty easy I think already. This is going to become much more awesome. I'm going to give it back to Alex now. You can talk a little bit where we are and how we've got to before. >> So who is behind OpenStreetView? This is the time actually so the guys up there are from Romania, it's an awesome time. I'm also from Cloosh, the guys from the bottom are mostly from Berlin. The guys below do the computer vision and the advanced stuff and. So we like to think of ourselves as a small startup. We work pretty isolated from the rest of the guys of the company. And, yeah, we're iterating fast and do pretty much a lot of start uppish things. What did we do so far? We started working on this about eight months ago building a very prototypish type of thing. Basically all the nice UI and the colors and all of that were added maybe two weeks ago or something. So as I said startup. And until now we collected over 4.5 million images and 140,000 kilometers. What is that? 80,000 something miles. So if you look at the program, the talk abstract, you'll see there are 30,000 new images and I don't know a couple of thousand -- no, 30,000 kilometers and maybe a million images. So just until we submitted the talk until now, we grew a lot. So this is our sample of coverage. This is just one act to better user. Better users. This is another. And another. Yeah, that's you actually. >> Uh-huh. >> I recently added this slide. This is San Francisco covered pretty well so far. Also parts of Europe like high level zoom of Europe. So, yeah. Martijn took care of really great program and some really active beta users, so I'll let him talk on that. >> Yes. Beta program, that went pretty well, we had great OSM mappers contribute tons of ideas and feedback during the past seven months probably, seven or eight months. It wasn't small, but it was like 50 people. Good number of them, fairly active. Using of very early builds that crashed a lot and still persisted. And I sent them weekly updates what we've been up to, we had weekly builds. And this beta program is going to continue. We're going to continue to kind of iterate fast with the beta users. And if you would like to be a part of that, you can get in touch with us. With me, and we'll make sure that you get on that too. Because that's really -- I think this beta program, especially because we had so many OSM mappers, I specifically reached out to a lot of people that I know. I think that made it -- that made -- that helped a lot make this into something that is really built for mappers, and I'm really happy with that result. So I'm also excited to continue that. Who knows what this acronym means? This was a thing on the Internet a little while ago. A day in the life of. You run out with your phone, and start with the OpenStreetView app, you plug in your dongle. This is optional, but -- Alex is going to talk about this stuff. It's pretty cool. So this is basically to enhance what we get back. It's a diagnostic dongle. So you have that. And this is one of my favorite dashboard mounts. Of course it needs to be on a dashboard if you're driving. We tested a lot of them -- there are a lot of crappy ones. A lot of crappy ones. There's some good ones. I like this one. I'll put the model name in the slide before I distribute it. So after a lot of testing, and, again, after a lot of feedback we got from this, we can recommend this particular model. So you stick it on your dashboard, start the recording mode, then you take it out of your destination. This is an older screenshot. Now it's a lot nicer. It wasn't back then. Uploading, as I said you do this when you have Wi-Fi, and you can do it on cell, but it would eat away at your data a lot. And then you see that the OpenStreetView website will share your tracks, you kind of browse through them. Use JOSM map, upload, be happy. That was quick, but Alex already talked about all the components. So a couple of thoughts; right? Some of you may remember there was a project already called OpenStreetView. There was. It was from quite some time ago. It was from 2009. John McCarol. Who of you know John McCarol? He's a super early OSM contributor, and he came up with this project in OpenStreetView in 2009, so we actually got in touch with him to continue kind of in the same vein. He had a project where you could upload images but basically video and images have other people use them to map. And there were actually 2,500 users on that budget, which is quite a bit of OSM project at that time. And there were I think some 80,000 images on there. And we were working with folks if they wanted to transfer their images to do OpenStreetView to do that. And like I said we were working with John to make sure that we continue in the same vein of openness and kind of in a close relationship to OSM. By the way, John McCarol is also responsible for this awesome piece of technology. This worked with -- this is called the where dial, and he basically makes social networks to -- you would put this in your home and if you go away and check in on four square or whatever, this thing would dial to where you would be at the stage. You could custom this. You could order custom -- I think you still can, but I don't think he likes to build them as much anymore. He has moved onto other things. I don't have one, unfortunately. So then the question who of you has used Mapillary? Yeah, so this question must be on our mind; right? So I'm going to talk a little bit about that. And I think -- so we like Mapillary, otherwise we wouldn't have done this either. And I think we can also do better for OSM by creating a project that is really 100% open in the sense component is open source. We don't really have any other idea or purpose with it than improving OSM. That's an important distinction. We also as a distinct you could say customized a little bit for the driving use case because our background is as a navigation-oriented company. You know, it can be used in other things as well. A little bit about that openness, I already mentioned this really. Everything is open source, on your data, you can call out any time. You can download your own data back out really easily as well. So basically uploaded it, but it really continues to be under your full control. On top of the software being open source as well. It is 100% OSM focused. That means that we don't have any other purpose with it than improving OSM. And we started -- like I've been involved personally with OSM for a decade now. We are all OSM fanatics, and we want to create something that hits home for OSM users and has that appeal to everyone and has that -- the power to compliment what mappers already do. And we did try to work with Mapillary first. We have scout signs that was a JOSM plugin. We wanted to get Mapillary signs in there as well. That worked for a little bit but then it ended up kind of being -- becoming a little bit pricey. So that's what kind of also triggered a little bit. So we can be a little bit more open I hope. So I would love to get your feedback on that as well. But think about that. That -- yeah, that place in the community that we like to have. To kind of close it, I want to give it to Alex to say a little bit about some of the technology stuff that we do behind the scenes. So, yeah, why don't you have a few minutes of that. >> So I'll be quick. Yeah, I'll talk a little bit about more of the advanced things that we do. And the first one is the computer vision technology where we do image recognition on street signs and also working on lane detection. So you can actually get a notification that -- a road sign was detected on the device, so you could get a pop-up, so we do the image recognition on the device, and then you have of course the map editor that Martijn talked about, and you can see all the signs, you can add it, you can add things and so on. The second part is the lane detection. So this is the development tool that's -- that we have. We're basically looking for the lane markings on the streets to detect, to detect the lanes. The other thing that Martijn briefly mentioned is the OBD integration. So what is OBD? It stands for onboard diagnostics for cars. Basically every car from the 90s until now has one. This is what they look like. It's usually somewhere in the left side of the driver's seat kind of hidden away. But the cool thing is that now you can plug in a Wi-Fi or Bluetooth adapter and then you have live car information on your phone. So what we're doing now is just connect to the dongle and to get the actual car -- speed of the car. So very precise speed. I know it doesn't sound like a lot, but I'll show you what we can do with it. So here's track from Berlin. In the left, you can see an image from OpenStreetView. By the way, this is not a good example of placing your phone in the car and also please clean your windshield, I'm not sure if you see but there's bird poop. [Laughter] Anyway so this is Berlin. You can see, like, in the right you see the blue, that's the GPS traced. But then with the red, it's our proved path. So what happened this guy entered the tunnel, the GPS got lost completely, you see the straight line. But just based on the accurate speed of the car and the phone sensors, we can calculate basically with a bit of algorithm magic there accurately follow the driver and you can see it on the map. So it's pretty cool technology. Yeah, so we support Wi-Fi and Bluetooth low energy dongles. It's integrated seamlessly in the app. And, yeah. Martijn, I think -- >> I'm going to -- yeah, this is it. You can still win one if you want one and go to our booth. Many of you have already been there. We're going to do a lot of stuff still. 360-degree support. The gifts are gone, unfortunately. You can download the app, though. Please do and try it out. Here's GitHub stuff for finding issues and I wish you all very, very happy mapping and safe trip home. [Applause] >> Sorry if I missed this earlier, but do you guys offer services that expose those images? >> You mean like an API or some sort of thing? >> Right. >> Not currently so we're definitely thinking about this. I very much want to. It's also a matter of scaling. Like, right now we need to find ways -- it's a little early on for that. It's definitely on the road map, though. Yeah. Good question. Thanks. >> I had a question about privacy. I saw a couple of the images there was a blurring tool and maybe some license plates were blurred. Is that a user thing, or are you doing that on your own? >> No. We're doing that automatically. So when you upload photos, we're first process them, remove faces basically blur them, faces and license plates. That's the law. >> To the OBD tool, there are a lot of Bluetooth dongles, Wi-Fi. >> If there's a list of application, of the dongles that are supported, we have a very small list right now basically this one that I show. So the problem is also a little bit the iPhone can support Bluetooth connectivity with these things. Because of Apple. And then there's -- the Android one can support Bluetooth LE, low energy Bluetooth and Wi-Fi. So we're going to hope to grow that list very soon. We're testing in a bunch of them, but it's not conclusive yet. So for now this is the one that we have in our booth and the only one that I can definitively recommend but that list is going to go up. >> One mention on that, we tested the top ten dongles from Amazon, and they all work. So pretty safe if you stick with those. >> So can you tell me about your relationship with Telenav and what they're getting out of their project? >> So we're navigation people, that's why we like the street view images. The project is Telenav project, that said, I feel that there's a sense of ownership needs to be from the community. So even though we're -- I feel like this is something we invested in heavily and that we're -- that we continued to invest in both in people and in money. But starting with this weekend where we release all of this stuff publically and put it out in the open. I'm hoping that we can get some sort of collected sense of ownership that really becomes something OSM rather than just something that we have. Does that answer your question a little bit? >> Sure. >> Okay. Talk to me later if you want to talk about that more. It's a good question. Thanks. >> A time for a couple more? >> I know you guys have a roadmap using computer vision for lane detection, and you guys thought about using it for business names and house numbers as well. >> We thought about it as well of course. It's something that you can do, no doubt. But I mean we have a lot of stuff to do faster so we'll get to that. >> Thank you for the presentation, a very exciting project. I think it's great there's more imagery. I'm seeing on your website that the data that you use is GPL and the pictures, how do you see the government project working, how do you -- like the collaborative around building something in OpenStreetView, how do you fit into the system of the -- >> Can you repeat the question? >> Yeah, I like this question a lot, and it's a good one. The eco -- how does it fit into the ecosystem? How do we get to build this together? So I think this is something we've been investing heavily in, and I think -- the best answer that I can give right now is we'll have to wait and see. I'm personally very committed, and I know our entire team is committed to making -- to bringing it out into the open and to support kind of the development as something that is also community guided that sounds very vague, I realize that. But I do want to make good on that. I think for all the team, it's very important to us that it is going to be something driven by all. I wish -- we're just out in the open it's going to be out in the open now but it's going to be a little bit wait and see, but I and the entire team are going to be here to support that, the development as well. That will have to do for now. >> I have a question about -- OpenStreetView is much easier than GPS mapping. But on iPhone or Android have open OpenStreetView much easier, much better, more of the same than -- >> Your question is if it's easier or that it is easier? >> OpenStreetView on your iPhone. GPS system is more responsive. >> Yeah. Sure in the app, I mean you could get started this afternoon. The app is free and there's no there's no restrictions of starting to use it right away and it doesn't really relate to -- directly to a GPS system in the car that you may have. So that's -- those things are really kind of quite separate. I would like to think it's easier to get started with, so I hope you try it and agree. Does that answer your question? >> I never used OpenStreetView before. >> Please try. Yeah. I would love to hear what you think. >> I was just curious if you put into thought into tagging images in terms of to say if you wanted to film or take photos out to the side of the vehicle out of the passenger window, does your system support that as metadata for that? >> Yeah. So we're recording a whole bunch of stuff including the heading of the phone so we can easily track the direction if you took it sideways or -- we actually tested with phones mounted in the rear windshield, and it worked quite okay. So. Yeah. >> I have a question about moderation in terms of do you have checks -- it seems like you have a lot of automated checks built in. Do you have checks to check malicious content or will that be human driven or moderators? Will you have a set of guidelines of what you can upload and can't upload? >> Yeah. I see that is something -- we have -- anyone can flag any image. On top of what we already do in terms of for privacy like the license plate, face blurring, we want to make it as easy as possible to of course delete your own images and flag others. What that process is going to be looking like is I think very much to all of us, I think, like, privacy is -- it's a very complex issue because it's not the same everywhere, even though the legalities might be the same, the perception of privacy may be still be different, and we can't -- I'm from Europe, I live in the U.S. now, I'm not sensitive of the whole spectrum of how people just of privacy anywhere. So I think it's important to have that in the hands of the users. How they're going to process that is something actively on our minds, but it's not yet -- it hasn't crystallized yet. It's a little early for that. But it is super important topic. So, yeah, again, that will have to do as an answer for now. All right. That's it. Thank you so much. [Applause]