Great read on the possible rationale - Summarising
#1 It Hurts Google - erodes googles dominance vis a vis commercial access to geospatial data + a viable reason to collect valuable location data on end users.
#2 It Complements Their Augmented Reality Business
#3 It Supports Facebook’s Place Data Generally
"what if street level imagery and supplementary data like reviews were built directly into Facebook’s products or at least resided somewhere on a Facebook property that could be linked out to? That at least would keep users in their scrupulously measured web of control."
I would flesh out one related/adjacent point.
~Google's mission statement: "Our company mission is to organize the world's information and make it universally accessible and useful."
~Facebooks Mission Statement: "give people the power to build community and bring the world closer together"
I'm not so naive as to suggest these ideals aren't deeply infiltrated/influenced by commercial imperatives but it is possible to see the way some of their behavior wraps around the stated principles. Google's credo of organizing information aligns with building out immense location data - and effectively mapping the world (and subsequently commercializing it). Whereas with facebook what I have noticed is - given their size and scale; they can follow their users (and subsequently envelope/"empower" user efforts to create community) ...if need via acquisition. In the digital world this has mean't buying the likes of instagram and whatsapp to ensure that community-building happens under the nexus of Facebook's control. If you accept that facebook believes in being where it's users are; understanding/being more proximate to users and how they traverse the real-world is invaluable - you are now more able to map out a digital proxy that is informed by person's experience of the real world. By creating a more informed digital proxy you embed a user more deeply in your ecosystem.
Facebook has completely failed. New York and Sydney have been thousands of miles apart since Facebook first set out to "bring the world closer together" and they haven't budged an inch since then.
But even if we dont take them literally: the united states is on the brink of civil war. We're more divided than ever, at the same time that individuals are spending record screen time engaged with facebook. Complete and utter failure to bring people together. The shareholders ought to have zuckerberg pilloried.
Its still puzzling to me why facebook wanted it. I have used mappilary a bit and it is not an alternative to google street view. The images tend to be shit quality at odd angles and with window reflections in the way and are not 360 degrees like googles are. The only purpose of mappilary in my view is for open street map mapping since the images are usually good enough to read points of interest on the side of roads so you can add them to OSM which street view does not allow.
More specifically, while it is probably legal to use Google Street View¹ and surmise facts about the world to map in OpenStreetMap, it is also a legal minefield and any contributions based on Google Street View put OSM at risk of costly litigation.
Facts are not copyrightable in most jurisdictions, so if you use Street View to glean facts (like names on signs or the presence of a playground), legally you are probably in the clear. But the OSM Foundation has no way of knowing whether this holds true in every jurisdiction without Google explicitly allowing for such use (which it won't, because OSM is a competitor of Google Maps), so use of Google Street View as a source is not valid in OSM because of internal guidelines.
For more information, this question and answer provide some insight:
>>Its still puzzling to me why facebook wanted it. I have used mappilary a bit and it is not an alternative to google street view. The images tend to be shit quality at odd angles and with window reflections in the way and are not 360 degrees like googles are.
Add FB's unlimited resources to it and see. What it is and what can be are different
I agree the quality of the images are very poor. Why go through all that effort of data collection with such inferior hardware? Most of the images are completely unusable.
"Unusable" for human eyes, or unusable for computers? The images do not need to be very pretty for a computer to extract street signs and other features.
I wouldn't all them completely unusable, they are just vastly inferior to googles. Personally I captured a bunch of images. If I was going somewhere already and the area was uncaptured I would just stick my phone up and capture the road. Then I would later go over the images and add things in to open street map.
My guess would have been AR being the biggest driving force. Oh, and every outdoor photo on Facebook can now more easily be “placed in context” on the map. Especially given Facebook’s “sell the users” business model, an accurate map of where the users are with respect to where the advertisers are would be quite useful to them (and it wouldn’t do to depend on Google, since they’ll butt heads as advertising platforms).
Making the platform free to commercial users as well might be more in line with Facebook’s business models — I’m guessing they’d rather create an API which allows them to exfiltrate end-user data, than charge companies for using their API.
> an accurate map of where the users are with respect to where the advertisers are
Not just advertisers - accurate location context information can (on it's own or in combination with other data) de-"anonymise" location data[1]. Any new sources of data that Facebook can correlate with their existing hoard of data increases the precision of their model of everyone's pattern-of-life.
> and every outdoor photo on Facebook can now more easily be “placed in context”
When you factor in their ML prowess [i] it's not hard to imagine that every photo you submit w̶i̶l̶l̶ is getting automagically tagged with an astonishing amount of context: who, what, when, and where (why too since they have all of your chats/messages too).
Wild prediction here, but I think this has largely to do with their VR/AR business. Sure, having a platform to manage more spatial data is cool, but I don't think this adds that much value to their current data.
In comparison, it could be bleeding edge for AR. There are few companies with a notable footprint in the street imagery business. Google is obviously number one, HERE has their own fleet and I think maybe TomTom and a few others actively capture this data.
What none of those companies have, however, is a flourishing VR/AR play. Zuckerberg is clearly bullish on VR/AR and Oculus is largely leading the consumer market.
This is where my prediction gets wild, Facebook intends to replicate real world environments in virtual reality. This is what OpenSfM could potentially do with enough hands on it.
I spent a considerable amount of time trying to do this myself. Photogrammetry is still in its infancy, but it works. I replicated my office in virtual reality using a phone app and Unity. I wanted to do the same with a single street, and that's how I first came across Mapillary. The technical challenge was too much for me, but I know it's possible based on literature.
Imagine doing this at scale with a network of vehicles and much improved software. You could, hypothetically, replicate the roads of New York City in a 3D environment. Imagine playing a video game in a real world environment, or simply being able to walk the streets of a place you've never been before. These could be transcendent experiences.
Your prediction isn't a prediction. It's fact. Just watch the videos posted from the previous Oculus Connect [1]. They demo some early prototypes to show their vision.
Several times I needed to pick my jaw off the floor from what they've accomplished with mapping the real world and virtualizing it. The fidelity is breathtaking.
If they keep up the pace of miniaturization we'll have Rainbow's End before we know it.
I'll maintain that it was a prediction as to why they purchased Mapillary.
But I'm super excited to see what comes of this technology. I recently bought into VR, and it's a clearly amazing technology. The content, however, remains lackluster. I firmly believe there is a large future in the production of VR content... To the point where I believe you could become a successful content creator with nothing more than a 360 camera and a few plane tickets.
Facebook need a sub CM accurate, self-updating, visual positioning map. Its the only practical way that "multiplayer" AR can work. (in fact any kind of useful AR)
SLAM only gets you so far (as we've seen with google and apple) as its only capable of tracking relative movements (if you have two phones, both their coordinate systems will be relative to their starting point, making accurate placing of assets impractical )
So what does mapillary do that helps?
1) the real time segmentation on a mobile phone is fucking spectacular, compare the speed and robustness to FB's panopticon, it blows everything out of the water
2) the geolocation of random imagery allows facebook to build out a sparse pointcloud suitable for generating cityscale visual positioning.
3) the anonymisation pipeline they have is also very good.
I tend to agree with this. Facebook also silently acquired the very amibitious UK startup called https://www.scape.io/ whose mission was VPS = Visual Positioning System. Their SDK/APIs allowed to build AR app easily, but now they're part of FB. That makes me think that its AR which is here at play.
Well worth a read. The article considers why facebook bought this. Maybe to help other competitors to facebook, cause they will be giving away all their mapping data for commercial use now.
#1 It Hurts Google - erodes googles dominance vis a vis commercial access to geospatial data + a viable reason to collect valuable location data on end users.
#2 It Complements Their Augmented Reality Business
#3 It Supports Facebook’s Place Data Generally "what if street level imagery and supplementary data like reviews were built directly into Facebook’s products or at least resided somewhere on a Facebook property that could be linked out to? That at least would keep users in their scrupulously measured web of control."
I would flesh out one related/adjacent point.
~Google's mission statement: "Our company mission is to organize the world's information and make it universally accessible and useful."
~Facebooks Mission Statement: "give people the power to build community and bring the world closer together"
I'm not so naive as to suggest these ideals aren't deeply infiltrated/influenced by commercial imperatives but it is possible to see the way some of their behavior wraps around the stated principles. Google's credo of organizing information aligns with building out immense location data - and effectively mapping the world (and subsequently commercializing it). Whereas with facebook what I have noticed is - given their size and scale; they can follow their users (and subsequently envelope/"empower" user efforts to create community) ...if need via acquisition. In the digital world this has mean't buying the likes of instagram and whatsapp to ensure that community-building happens under the nexus of Facebook's control. If you accept that facebook believes in being where it's users are; understanding/being more proximate to users and how they traverse the real-world is invaluable - you are now more able to map out a digital proxy that is informed by person's experience of the real world. By creating a more informed digital proxy you embed a user more deeply in your ecosystem.