Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Project Tango

google.com

Post reply on HN
Print view View original

231–240 of 353 posts

Re: Project Tango

#231

Post by boucherm »

gallamine wrote: post #29
gamegoblin wrote: post #17

I wonder if it uses the machine vision aspect to prevent the typical problem of accelerometer drift. e.g. by orienting itself relative to walls/other stationary things. Also, imagine making a 3D "scanner" that you can scan objects with into a virtual world, or print out on a 3D printer.

Accelerometers usually don't have significant drift - digital gyroscopes do. The acceleromters are quite noisy though (and measure both gravity and acceleration). From some of the brief images in the intro movie it looks like the cameras are doing tracking and alignment of "interesting" features. Things like that fall into the Simultaneous Localization and Mapping (SLAM) technology area. Very cool stuff.

Well, maybe was I just unlucky, but I tried to use an IMU for my SLAM and was shocked by the extravagant amount of drift of accelerometers.

Re: Project Tango

#232

Post by agentS »

hrjet wrote: post #226
cbr wrote: post #191

Earlier quoted context omitted.

The solution is to stop centralized data collection. Centralized data collection gives me good auto-complete in search. It means when I start sending an email to John Smith and Jane Doe it can ask me if I might have meant Jane Roe instead. It means when I lose my phone I don't lose any of my data.

A decentralized solution for that would be to have user's data stored in user's private cloud, encrypted and cached on the client.

One of the reasons that spelling corrections are so good on Google is (probably) that they are machine learning models trained on query logs(1). i.e. if you search for "hacer news", and then without clicking on any results, issue another query for "hacker news" in a very short timeframe, then it will learn that "hacker" is a good suggestion for "hacer"(2).

Similarly for "Mark as Spam", Priority Inbox, Recommended Videos on Youtube, Voice Recognition on Android, etc.

Note 1: Yes, you could also do a pretty good job by having a model of your problem. i.e. computing a weighted levenstein distance where the weights are the probabilities of making that error. However, I'd argue that this would still be better with centralized data; you can compute much better probability vectors. And regardless, the best solutions in the field will be with the combination of both.

Note 2: All of the above is speculation. While I help write some of the tools that these guys use, I have no knowledge of how they write their software. This is just how I'd do it.

Re: Project Tango

#233

Post by ulyssesgrant »

siera wrote: post #221
ulyssesgrant wrote: post #219

As someone who's worked on visual odometry projects before, I'm extremely interested in how they're able to compute the phone's motion estimations so accurately and in real time on a mobile device. Is any of this project open source? Is anyone here familiar with the vision techniques they're using to accomplish this feat? Perhaps there's an FPGA or ASIC designed for visual odometry inside the phone.

It uses a Myriad 1 processor from Movidius http://www.movidius.com/tango . It's a custom media coprocessor aimed at mobile devices. I found more information about the processor architecture in these slides : http://www.hotchips.org/wp-content/uploads/hc_archives/hc23/... Edit : Video of the talk associated with the slides http://www.youtube.com/watch?v=KiTaPRtN2yM

Thanks for the info. Still very impressed with the speed and accuracy, assuming it works as well as they have presented. To be able to compute visual odometry in an arbitrary scene on a mobile device seems like a really big accomplishment, no?

Re: Project Tango

#235

Post by torbit »

evanlivingston wrote: post #81
martythemaniak wrote: post #27

This is essentially putting a Kinect in your phone and hooking it up with hopefully high-level APIs. It may take 2-3 years to make it into regular phones, but when it does it will be huge. Apple's acquisition of PrimeSense (the makers of the 1st gen Kinect) means they're also working on this. To give you a real-world example: when I started BarSsense ( http://www.barsense.com ) the core problem was tracking the path…

I don't understand the application beyond gaming. The other applications discussed are basically shopping. Essentially this would make it much easier to represent the physical world digitally. But what use cases does a consumer or the average phone user have for digital representations of the physical space around them, particularly given that the user is already aware of the the physical space around them? How can t…

wow did you post this on the wrong website. "I don't understand"

Re: Project Tango

#236

Post by wwweston »

evanlivingston wrote: post #81
martythemaniak wrote: post #27

This is essentially putting a Kinect in your phone and hooking it up with hopefully high-level APIs. It may take 2-3 years to make it into regular phones, but when it does it will be huge. Apple's acquisition of PrimeSense (the makers of the 1st gen Kinect) means they're also working on this. To give you a real-world example: when I started BarSsense ( http://www.barsense.com ) the core problem was tracking the path…

I don't understand the application beyond gaming. The other applications discussed are basically shopping. Essentially this would make it much easier to represent the physical world digitally. But what use cases does a consumer or the average phone user have for digital representations of the physical space around them, particularly given that the user is already aware of the the physical space around them? How can t…

> But what use cases does a consumer or the average phone user have for digital representations of the physical space around them, particularly given that the user is already aware of the the physical space around them?

For the user? Probably not much at the moment, if ever.

For a company whose mission is to collect, aggregate, and extract monetary value from every last piece of data in the world?

For a government that's interested in extending its awareness?

Priceless.

Re: Project Tango

#237

Post by aestra »

Gigablah wrote: post #199
evanlivingston wrote: post #115

Earlier quoted context omitted.

I think I'm a neo-luddite.

This has applications even for, uh, "normal" people. For example, online real estate portals. People who are house-hunting would probably appreciate floorplans or a 3D tour. Agents, landlords or property developers would be able to easily create those with their mobile device.

>Agents, landlords or property developers would be able to easily create those with their mobile device

But they won't! It is very difficult to get those people to post more than a low res picture of one room online when they can very easily walk through the house with a video camera, for example. Even large apartment complexes have 1-2 pictures on their website and call it a day.

Re: Project Tango

#238

Post by girvo »

jrockway wrote: post #225
gilgoomesh wrote: post #193

Earlier quoted context omitted.

I don't think you're a luddite. With any technology like this, it's important to ask: what applications does this really enable and will those applications simply be a waste of time. This is an unveiling of a technology but the "applications" they show in this video are probably a waste of time to most users. Google have not shown a killer app that uses this tech. Maybe they're working on something (they hint that th…

Google Maps already has indoor maps.

Yeah but they even said in the video that they struggle with indoor navigation and positioning.

Re: Project Tango

#239

Post by discostrings »

moultano wrote: post #229
discostrings wrote: post #222

Earlier quoted context omitted.

> This is a very myopic view that has already caused many deaths. Privacy concerns around medical records have prevented any central database of them from being built, and thus prevented us from saving all of the lives that could be saved by analyzing that data. This line of reasoning seems similar to arguing that, by not creating a society in which 95% of people live in sanitized hospital rooms and the remaining 5%…

It's a cost-benefit tradeoff like anything else. Unfortunately, the costs are filtered through fear-mongering and the benefits are trade secrets, so the debate is particularly ineffectual. The thing I find most aggravating about it is that the standard for harm for data seems to be "What could a totalitarian government do with it?" and there are very few useful things that couldn't be used for very bad things in the…

Part of the problem is that privacy, at least in this context, is framed as a "debate" at all, rather than as a matter for individual choice.

Some people would gladly share medical information for the greater good if asked. You'd have to pry it from the cold, dead hands of others. The problem is that we currently don't separate those who would like to provide their information from those who would like to refuse because we're greedy for as much information as we can get, or we don't trust people to make the "right" decision.

A system that worked well and that accomodated both views would be one that really respected a user's choice--one that could handle data in a centralized or individual fashion, accordingly. Data that was willingly given could be used under the terms of the agreement without risk of angering people who don't want their data to be analyzed. It would ultimately provide the same benefits while respecting individuals and not creating controversy.

In the medical records situation, you'd have a group of people at one end that would be quite pleased to contribute their information, a group at the other end that would immediately decline, and quite a few people in the middle who would likely fall somewhat evenly to either side. Sure, some people would probably be fear-mongered out of sharing, but I imagine there'd be a lot less of that going on if people knew they could actually make a meaningful individual choice for privacy if they chose to do so. When opt-outs are buried, hidden, and it's not clear that they actually work, skepticism and fear grow. In a system that doesn't try to railroad people to sacrifice their privacy--a system that actually respects the individual's choice--fear would be reduced.

The only way to cast questions around this type of privacy as "a debate" involves changing the issue to compelled sharing of personal information. In a debate over forced participation, I think it's pretty easy to see why thoughts start to drift towards totalitarian concerns.

I think a system that offered real choice for each individual may drastically reduce the current problems from both perspectives, at the price of some engineering overhead.

Re: Project Tango

#240

Post by brymaster »

blhack wrote: post #165
SiVal wrote: post #53

"Imagine measuring your room by just walking around in it." Yes, and imagine transmitting each of those footsteps in real time back to Google, so they can have a map of your room, too, in case they, or other parties, ever need it. I was just having a discussion yesterday with a friend who works at Google about what data they store when you query their search engine. Every single keystroke, including backspaces, is st…

Good god I am so sick of reading comments like this. This view of the world is just lazy . Almost as lazy as "remember, if you're using a service for free, you're not the customer, YOU'RE THE PRODUCT!", which seems to get parroted every single time there is any discussion related to google, facebook, microsoft, instagram, twitter, or any other web company that uses ads as a revenue model. Yes, google can know what yo…

> Good god I am so sick of reading comments like this.

Then don't.

You just lazily suggested to these people to stop using these products/services, so I suggest for you to stop reading comments "like this."

> Almost as lazy as "remember, if you're using a service for free, you're not the customer, YOU'RE THE PRODUCT!"

By the way, what do you think Facebook bought yesterday for $16 billion? Not a product, but users.

Post reply on HN