Hacker News new | past | comments | ask | show | jobs | submit login
Introducing OpenStreetView (openstreetmap.org)
455 points by progval on Aug 21, 2016 | hide | past | web | favorite | 79 comments

Similar to http://mapillary.com.

It's great to see more open data like this, lots of uses for creating maps by hand or even machine learning.

I haven't used mapillary, so just a question: is the user-generated content made available to the public?

To me it seems that OpenStreetView was made especially as an answer for Mapillary. Mapillary has almost 80 million photos uploaded by enthusiasts, who did it free charge.

Yet, even for a simple embed on a website, you have to pay them $99 / month. Using the images for custom maps: $249 / month.

Accessing the data is not publicly available, but is probably in the range of enterprise pricing options, only listed as "Contact Us". To me this seems to be the opposite of the OSM model and there is a huge need for a OSV like platform.

Exactly. OSV is totally about being Open and free (as in both freedom and price).

Can OSV/OSM contribute the same resources to their project as Mapillary is contributing to their own?

Mapillary is VC-backed. OSV is backed by Telenav, a publicly traded location services company. Your call as to which is the more sustainable model!

(The OSM Foundation has no connection with either project and is traditionally focused on its core task of collecting and distributing openly-licensed vector geodata.)

Is it free to use OSV images and data in commercial projects?

Images are openly licensed. Data licence hasn't been published yet but the project creators are making encouraging noises (e.g. in the comments thread of the linked post).

There's back and forth about that in the comments at the link. Note that 'jesolem' over there is the CEO of Mapillary.

The short answer is that if you sign up for a normal account and upload a photo to Mapillary, it appears for all other users on their website and is available in some OSM editors. Getting the complete raw data is not currently possible.

Mapillary also apparently offers private databases to users, those photos aren't available to the public. But those users aren't going to be naive to that fact, they are paying for the private instance.

Mapillary offers free bulk access to the imagery, but with the catch that the location metadata is excluded(!). This makes the bulk image access effectively pointless. OpenStreetView seems to have every intent of opening bulk access to both the imagery and the metadata; which will finally make it useful.

You can get that metadata through the APIs.

Very cool. OSM has been the default map on my phone for years. I do worry that the name is a little close to Google's trademarked Street View (see list of Google TMs here: https://www.google.co.uk/permissions/trademark/trademark-lis... ). They might wish to avoid an unnecessary lawsuit by calling it something else.

The name should be OK since there has already been an OpenStreetView since 2009.


Which could also be a name Google objects but just hasn't found out about yet.

Street View itself has been around since 2007.

"Street view", at least in English speaking countries is a generic name for a street view service. It's most probably not a valid trademark alone (but Google StreetView might be.)

> a generic name for a street view service

What is "a street view service" other than Google Street View?

Bing has a similar service called Streetside. I can't find any usage of the phrase to mean anything other than Google's service.

Mapillary uses the phrase "street level photos"

It is a registered trademark as well: http://tmsearch.uspto.gov/bin/showfield?f=doc&state=4805:dq9...

I think in a dispute, Google would very likely win.

> I think in a dispute, Google would very likely win.

In the US. But OSM doesn’t have to care about that – the worst that could happen would be them being banned from doing business in the US, or their profits being seized.

Which is no risk to them.

OpenStreetView isn't directly associated with OpenStreetMap or the OSM Foundation.

It's run by the US company Telenav (all over the terms: http://openstreetview.org/terms/ ).

I guess individual users are giving OpenStreetView/Telenav some permissions via OpenStreetMap oauth.

No permissions per se, we just didn't want to have users create YAA. We figured if you're going to be interested in OSV you'd already have an OSM account anyway.

That needs to be fixed, OSV should be owned and run by the OSM folks.

The OS mapping agency in the UK has been using "Street View" as a brand name for over 10 years. Google only use "Google street view" I think.

I understand that it's a volunteer effort, and volunteers don't have the hardware required for even 180 degree view, but most pictures are taken out of the front of the car, limiting its worth (2/3 of the picture is road, not the houses on the side).

It would be more useful if people stuck two phones (one on each side of the car) to take pictures of the passing buildings rather than road

When it comes to "useful", the important question is "useful to _whom_". To a mapper, even a single photo well geocoded is a valuable source of information and its presence on a system like OpenStreetView (or even Mapillary) is extremely useful as a record of the original source.

Seeing as, for now at least, you need to have an OpenStreetMap account to contribute to OpenStreetView, it's pretty clear that the initial target market for usefulness is mappers.

Wonder whether you could use something like this:


Or whether there are any affordable 360 degree cameras which exist.

Yes - there's the LG 360 Cam, which I've got and which is (generally) a lovely bit of kit. The Ricoh Theta is similar though more expensive, and Samsung do the Gear 360 but it's less pocketable.

We're in the process of adding 360 camera support.

I've got a Theta S and it's a wonderful, simple piece of kit. Obviously one of the drawbacks of most handheld 360 devices in this context is you'll be in the shot, unless you mount it out of view and use the app to take the picture - which is a bit of a faff to ensure on the fly.

Some phones come with very good "wide-angle" cameras. Not 180, but similar to what a GoPro gives. Ever since I got one, I've been amazed at how I ever got by with a crappy "boxed-in" field of view camera on a phone.

With the wide-angle view, even pointing it forwards you can get quite a bit of the surrounding area.

Exciting! Now instead of waiting for the street view car to pass through your neighborhood, you can add the pictures yourself!

I wonder how privacy issues are handled though. Is any blurring of car numbers or faces being done or planned?

If you check the presentation, there is a question about this and they say that they support it. Here it is https://youtu.be/XuGjahx8CFc?t=27m32s

It is, but I've seen many unblurred faces. On this one the car wheel on the left is blurred out though: http://openstreetview.org/details/10510/4664.

Why blur faces when it should be possible to completely remove people? At least eventually.

It's a lot easier to detect and blur faces than to detect people, remove them and then try and fill in what was removed.

Now I wonder if phones have been made with some kind of z-sensing to provide depth mapping for the image. You could share your data and cross corelate like Microsoft picture merging technology.

Google's Tango project is working on depth sensing. They don't expect (in a UX sense) it to work over longer distances because of power requirements.


Right; I forgot this project. Can IR be power efficient for city/boulevard scale depth ?

I think power will always be an issue for something built into a phone. If it could run off the car, the power considerations would likely be irrelevant.

I guess nothing much will happen with it, a patchwork of point clouds is less interesting than a patchwork of photos.

Mapillary has done some work on recovering a point cloud just from correspondence between the photos.


Sure, handhelds have bounded power. I don't know how depth sensing tech work but I expect it to be power hungry. Question is can it be LCD/WiFi demanding or would it kill the average battery after 10 shots.

To a first approximation, it's the same kind of power as using the camera flash with every shot. It's just a "flash" that's not in the visible spectrum.

So for a car-mounted system surveying streets, the power would be comparable to having headlights on all the time.

>Is any blurring of car numbers or faces being done or planned?

Why would that be necessary? All the pictures were taken in public so there's no expectation of privacy.

That is a very American-centric privacy perspective and does not apply to most of the world.

Freedom of panorama is quite widespread: https://en.wikipedia.org/wiki/Freedom_of_panorama (And it is expected that it will be part of the harmonized data protection law that the EU is currently planning.)

Also, at least in Germany, the right to your own image does not apply when "persons only appear as props next to a landscape or other location". (Whatever that means is usually decided by the judge AFAIK.) Source: https://www.gesetze-im-internet.de/kunsturhg/__23.html

The law in your second link explicitly spells out that this is only valid as long as there is no "justified interest" of one of the people visible that the image isn't published. Which is a pretty serious limitation, and basically means you have to review and judge every image individually. In practice it's explicitly not "happened in public, no expectation of privacy".

If you pay some attention you're probably going to be ok, but the "probably" is the annoying thing for individual actors. (E.g. I think Google lost a few lawsuits in Europe and Canada due to people being recognizable even with their face blurred. Annoying and expensive for Google, a criminal conviction and really expensive if it successfully hits an individual.)

Good thing for an open, collaborative project is that you are not Google and can be more aggressive with filtering questionable images. Google seemed to value having complete coverage of areas quite a bit, open projects don't have to as much. Some data is better than none, and holes can be closed later, step by step.

Freedom of panorama is a term from copyright, rather than data protection. It's primarily about designed buildings - are you allowed to make/share a photo of a building where the architect (!) does not allow it... It sounds absurd like this, but the arguments against it are not that easily ignored. Public interest should outweigh it though. At EU level the issue is that the freedom of panorama does explicitly not exist in some countries, now the laws get a bit more harmonised at EU level and the new rules may/may not force some countries to re establish freedom of panorama. Let's hope the European Commission puts it's foot down and actually makes it obligatory, but chances are that whatever the commission proposes will be ripped apart when member states have to discuss & find a compromise.

Even in America, it hardly seems universal in the social sphere; Glassholes didn't take their nickname from spying on private property.

start recording people in public and some percentage of time you are going to get assaulted. Multiply that percentage by 10 for police officers.

People have a reasonable expectation that their personal data (their face, their car number plates) are not going to be gathered and processed without their knowledge or permission, and that if it is going to be gathered and processed it's going to be proportionate to the need.

Unblurring is likely to be legally problematic in EU.

That's not as simple in many parts of the world. Hiding faces is a good starting point to avoid trouble with local laws.

> All the pictures were taken in public so there's no expectation of privacy

Just because you can go around publicising random people's locations, doesn't mean you should.

This is pretty cool.

What would be amazing to also have LIDAR sensors (EXPENSIVE! I know, I know) which would really make amazing data for self driving/completely autonomous vehicles; not only cars but also drones, etc.

Anyone know the pro/con for implementing LIDAR (although it maybe super late before we have enough data for effective navigation at long distances), and if it is worth it?

I think it would be really cool to have info about the heights of buildings, exact positions of traffic lights, etc. which are miniscule but crucial for low altitude drone navigation. Or can we infer almost everything from the picture data?

How can you ensure the data remains useful in the face of construction, tree growth, and other changes? I think if we are to allow drones to fly unsupervised, they will need to be limited to annotated environments (that is, locally broadcasting navigation beacons) and will need full sensor suites anyway.

Well the mapping frequency must be increased to account for stuff moving around, but the data like height of buildings is not going to change a lot. But I do get your point of the impracticality.

Autonomous self-driving cars are hopefully going to cause the price of LIDAR sensors to go down. Once they get super cheap, sure, load up 10 or 20 LIDAR sensors on anything with wheels.

Yeah I'm hoping for that. LIDAR has a lot of useful applications not only outdoors but also indoors that the average hobbyist can put to good use.

Perhaps something like this instead? http://hforsten.com/homemade-synthetic-aperture-radar.html It would still give some interesting information.

That looks super interesting. Thank you!

Interesting that some roads in Germany are covered, at least for now.

It will be interesting to see if that remains. I suspect that a lot of the privacy objections around Google Street View are objections to google, not objections to street view.

The few I randomly clicked on seemed to be either tests without outside view or with mostly-blurred faces and license plates (some unblurred, I suspect where the face-detection failed), so there are efforts to conform with privacy demands here.

Google Street view also blurs faces and license plates. What killed it in Germany was that some people wanted to be able to have their houses blurred out, and processing these requests took lots of manual work.

The big issue was that Google had the cameras above eye height – German law only allows panoramic images at eye level.

Google's fix for that should've been to genetically engineer a 3-meter tall man. It's eye level for him!

Germany has a history of "Let's make a new race of super humans, and presume they are the default". So I don't think that argument will fly.

I vaguely remember websites dedicated to providing HD images of all the buildings that are blurred in Google Streetview, as another instance of the Streisand Effect.

You can have your house blurred out in the USA, too.

Didn't the courts require that people be allowed to blur their house fronts? Which resulted in large amounts of Google Street View being blurred.

Why? Google Streetview stopped because public backlash, not because there's any law against it (it's actually even explicitly allowed through Panoramafreiheit)

Actually, Panoramafreiheit does NOT apply to Streetview, because the cameras were not at eye level, and could look into gardens, over fences and hedges.

This was tried in court, and Google failed.

Panoramafreiheit only applies at eye level (I’ve talked with a few people studying law about this due to the current drone debate)

Props to the people who made this - fantastic public contribution. My big wish list as someone who works in urban design and planning would be this paired with 3d point cloud detection to make a 3d model of streets and spaces. Maybe one day eh!

As long as data is available to download by regions, streets etc and combined data is freely available it will be possible.

You can do it already with Flickr. They provide metadata of images, you can selected images from a map/region.


Mapillary already does that?

Some of the images appear to be upside down. Is this a known bug?

It doesn't seem to be in their tracker. Mapillary had a similar problem, it turns out that the camera API was updated to provide the orientation sensor orientation and they were using the older API on a phone with a rotated sensor.




Yes, most all of these were produced with an early beta of the apps. Has since been fixed. There's still a known issue where images appear mirrored in OSV when using certain Android phones, https://github.com/openstreetview/android/issues/3

Every huge entity having their own mapping system in order to grind out a minor efficiency is probably the biggest tragedy, to me, in modern services. Before mapping, it would be the TLS certificate racket, which LetsEncrypt appears to have dead to rights.

What I've wondered for a while is the feasibility of producing streetview imagery with drones. It seems to be the perfect solution for distributed street shooting.

Anyone know what the business models of Telenav and OSV are?

No idea, but a guess: OSV will improve OpenStreetMap, which ensures that there is more options for map data?

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact