Hacker News new | past | comments | ask | show | jobs | submit login
Early Days on Street View (ambivalentengineer.blogspot.com)
204 points by mholt on July 3, 2014 | hide | past | web | favorite | 32 comments

Pages Jaunes (Yellow Pages France) had street view in March 2002. I used it to work out how to get from the train station to accommodation on arrival. Two problems though on arrival train partial strike so travel agent arranged a car and Page Jaunes assumed a side exit of Les Halles railway station was the main one. Still earlier than 2005, and to quote archive.org "2 200 000 photos de villes" https://web.archive.org/web/20020326031423/http://photos.pag... and voila! https://web.archive.org/web/20061130215611/http://photos.pag... and https://web.archive.org/web/20051012124517/http://photos.pag...

I remember this, but it was very different from Street View; it was made by sending students on rollerblades to snap pictures of buildings; it only concerned itself with big cities (mostly Paris at first if I remember correctly); and on the client side there was no stitching of any sort: you could just see a picture, then another, then another.

The idea was there, but the execution was poor; execution is what matters.

So you could actually walk down the street like SV? In 2002? Wow.

I went through old emails and I was using it in early February 2003, so not quite 2002. I showed it to a number of people and some said, well, it's interesting but will never take off, others thought it looked really useful. One comment "Hey there is a boulangerie (bakery) right where you are staying!", this is cool. (Pity the bakery went on holidays just as we arrived). As others pointed out it was buildings only, so I knew what to expect to my left and right, but had, for example no idea how wide Boulevard de Sebastapol was but I did know there is a KFC on the left (still is looking at Streetview now). One good thing about the pagesjaunes solution was that they walked rather than drove the streets. As my journey was along Rue Berger (from Chatelet les Halles heading towards the Pompidou Centre), the 2014 Streetview experience is pretty hit and miss.

Still I thought it was cool, and even took a bunch of photos (Digital, brand new Sony Mavica writing to CDRs in the camera) of the walk from the station, duplicating what I saw before setting out.

I hate to be a critic, but to me the remarkable thing with Google's Street View is how stagnant the interface has been. They have collected so much data, but using it is still so painful, really a last resort. It is impossibly slow to explore an area. Heaven forbid you click on the wrong side of an intersection and have to click a half dozen times to cross the street. There has to be a better way.

I am not sure how much they have rolled out, but Microsoft has been much more innovative in presenting Street-View-style imagery, e.g., using linear panoramas along a street instead of a series of spherical panoramas.

This. Google has so much data at this point they should be actively converting it all to pure 3D (structure from motion etc). With tango on the horizon there will be a point in time phone could know where its at by simply comparing geometry, or at the very least augment camera view with accurate directions based on the scene geometry, and not crude 'somewhere this way' gps+magnetometer Arrow. They could also use this data to augment self driving cars, so its a no-brainer.

Google Earth is PATHETIC right now, sparse hand made badly textured blocky models of few selected buildings in a capital of European country? Really Google?

I sure hope they have a group inside google working on this already.

A UI ripped from video games would be a big improvement for many: WASD + mouse, for example.

What was the difficulty in making the camera unit watertight? I'm guessing a simple solution like transparent globe would cause reflection problems but why couldn't each camera be housed in a rubber-tight casing? The lens could be made waterproof by design and everything else would be in a sealed rubber glove. I'm not implying that the guys who mapped the entire world in 3D couldn't figure out how to waterproof cameras. I'm asking what the core engineering problem was in attempting to waterproof these cameras.

First you need the thing to be optically clear. Then you need it to be watertight even with 100 MPH winds blowing on it. And once that's done, you'll find water inside anyway. Condensation can easily form in a closed container which moves at varying speeds through varying temperatures and has water cooling it suddenly sometimes. You may even end up needing a heater inside the thing before you've got it perfect.

WiFi packet payloads:

> So in theory, we should have just grabbed the header. However, a single wrong line in the configuration file caused us to grab the packet payload too. Had we ever audited the data on the disks, we would have found the bug and fixed it, but... nobody ever noticed. Oops.

That is Google's official opinion of the story.

Some european countries (Germany, Austria) bag to differ. Google captured all WiFi data (unencrypted & encrypted). As a result of the war-driving (http://en.wikipedia.org/wiki/War_driving ) Google StreetView cars were banned in these countries (Google had to delete the harddrives) and as a result there is a huge empty spot on the StreetView coverage in central Europe (only a few cities in Germany, no street coverage at all in Austria): http://en.wikipedia.org/wiki/Google_Street_View_in_Europe#Co... . Some attacked the car: http://www.austriantimes.at/news/General_News/2010-04-08/223...

one of many news articles about this issue on heise.de (German): http://www.heise.de/newsticker/meldung/Datenschuetzer-Street...

Google promised to delete the captured WiFi data but failed to do so: http://www.telegraph.co.uk/technology/google/9432518/Google-... , official letter: http://www.telegraph.co.uk/technology/google/9432552/Google-...

I have a hard time believing that Google deliberately set out to capture wifi payloads. Apart from the obvious illegality, what use would they possibly have for such data? Google already knows an immense amount of detail about its users.

They screwed up (twice, if you count failing to delete the data) and thats a serious matter. But deliberately wardriving entire countries? Hard to believe.

I think there's a middle ground here - first off, is it that obviously illegal? Is there an expectation of privacy, similar to eavesdropping on a conversation at the table next to you?

I think it is now, but if we go back to 2007/2008, it might not have been.

And it might not even have been malicious - the Google engineers doing it might have said, "Hey look, if x amount of wifi data is good, X amount has got to be even better!". After all, they're data junkies, and might have decided to capture chunks of data wholesale with the best intentions (and maybe not even any clear ideas of how to use it).

I'm not trying to justify their actions, just trying to posit a scenario where Google did intentionally capture wifi payloads without requiring that they had a malicious purpose to do so.

Anyone have a mirror, it looks like Google was not happy about this being discusssed:

> Down until I'm done talking to Google about this post.

Wow, I didn't realize that Amazon Blockview (part of A9 at the time) was so short-lived. I used it all the time in Seattle when it was running.

Why didn't the team just bolt together off-the-shelf DSLRs for this?

It seems like a lot of the low level problems were already solved by commercial camera makers many years before.

Even today with the R7, why not just bolt together Gopros and manually set the Aperture, ISO and shutter speed. You'd get better quality images, waterproof cameras and they'd be significantly cheaper.

It's a shame to see a talented engineer distort the truth about Wifi packet collection, especially after Google were shown to have acted deliberatly (rather than "a single wrong line in the configuration file") source: http://www.independent.co.uk/life-style/gadgets-and-tech/new...

The beginning of the article mentions using off the shelf parts, with terrible results. I'm sure if it was that easy and they got decent results from it, they would not have expended so many resources custom building stuff.

If they are still using the same R7 design since 2010, it seems like they've decided that the current hardware and image quality is good enough, and that they would rather focus on expanding the data set.

I'd be interested in an open source solution for displaying our own streetview-type imagery. The county I live in is quite rural so Google hasn't yet driven our streets. In 2011 the County's public works department paid a consultant to drive the public roads in a streetview-type car. The county has all collected imagery and point clouds, but only outdated software to view it in. I've hacked together a proof-of-concept that loads the panoramic images into a web browser (using three.js) to pan around and zoom. But I haven't figured out how to overlay the road centerlines yet to be more Google-like.

> "I went on to develop aerial cameras for Google, but that story will have to wait for Google to publish some details."

I'm wondering what, if anything, that statement has to do with the Skybox acquisition, or if the two are previously-unrelated parts of an overarching aerial imaging plan.

Google is so far ahead of others in mapping it is downright scary.

I read somewhere speculation that Apple's iBeacon could be used to create precise indoor maps of public or semi-public spaces. But then I read about Project Tango, and wonder...

I suspect those aerial cameras are for airplanes; it's been known for a while that Google Maps doesn't just use satellites.

Or balloons.

you guys are so deep in google reality distortion that you probably never checked out microsoft maps.

it has what they call birds eye view

it is a mix from street view and satelite. pictures taken form low flying planes at a angle. and it can be rotated.

you can get a sense of the neighborhood location you are looking at AND see the buildings on the street.

i saw it the first time by mistake. i was also in google reality distortion field. when i got a hp touchpad to install android, but while using the palmos maps, i was greeted with that pleasant surprise.

The thing is they buy these photos from outside vendors and I imagine they are very expensive to get. There's decent coverage from urban areas, but I don't think it scales up as well as street view driving. Also, it's certainly neat for checking out areas in general fashion, but I'm not sure it's especially useful.

EDIT: predicate

funny. you said the exact same thing people said about satelite/street view, before they covered most bigger cities.

Ahh yes, the ladybug, we had a few of those back in 2005-6. One of my colleagues strapped one to the top of a car with a GPS and drove around downtown. We wrote a java applet (and eventually had a flash version) to navigate the capture sequences, eventually integrated with Google Maps. Someone also strapped it to a robot and did our school's campus (or at least as far as the battery would go). Good times.

Great story of a fantastic project; Street View is still the main difference between Google Maps and other alternatives (Bing has "Streetside" but with a much smaller coverage; and AFAIK OSM / Mapbox have nothing).

What an excitement it must have been to work on such a project! Thanks for sharing.

> and AFAIK OSM [has] nothing

Some of us are working on fixing that.

Mapillary is doing crowd sourced street imagery: http://www.mapillary.com/map

But they, as far as I know, are not connected to Mapbox or to the OpenStreetMap project (Your phrasing groups those two things more than I would).

Here (the former Navteq, part of the Nokia that isn't part of Microsoft) is probably running the second biggest street level photography project.

There are a few more comments by the author on Google+: https://plus.google.com/+StephenShankland/posts/GGQdgLPn8J5

The things they were doing with cameras reminded me how Android started out as a camera project and not a smart phone. I wonder if there's any relation...

Registration is open for Startup School 2019. Classes start July 22nd.

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