Hacker News Reader

Comment by npunt | Hacker News Reader

Oy. This is how not to launch new tech - they're announcing hardware without any use cases to explain to consumers why it's valuable. This is a symptom of the lack of coordination between electronics manufacturers and the Android team, which unfortunately doesn't seem to be righting itself after all this time.

I love 3d mapping tech, and am eagerly awaiting it in devices. This tech could have been integrated into Google Maps, one of the most popular apps on mobile devices, but instead is launched with this weird message of hope that maybe one day it could be used for useful things potentially, you know, if developers come up with some ideas for that or something. This is Microsoft all over again.

Techies of course can speculate and hope and fill in some possible use cases, and it'll maybe move a few units for a short while as a lot of gimmick tech does, but to the general market this is one big 'huh?'

Use cases, folks. Use cases.

View on HN

Well, for what it's worth we are actually doing the best use case of this (AR home furnishings design) with inside out 6DOF markerless tracking on iPhone and iPad without needing the Tango system.

Monocular visual odometry is becoming good enough that depth cameras aren't worth the overhead. Samsung even asked me about a year ago if it was worth doing and I told them no.

Can your system make out where the walls and other objects are in the room in real time, and thus make sure virtual objects do not overlap?

One striking demo of Tango, back when they unveiled their dev platform tablet, was that you could use the tablet to "throw" virtual balls into a random waste basket put on top of a chair. Everything computed in real time based on what the sensors picked up about the world around it.

So if you want to measure up your apartment, all you basically have to do is take a Tango device for a walk around it.

You are basically asking about a few different things - though they are related.

Can your system make out where the walls and other objects are in the room in real time,

The first thing is plane detection and tracking. The answer is somewhat. We can detect and track some planes, but it's not yet up to par with native depth sensing systems in ideal environments. It performs better in high sunlight environments however.

thus make sure virtual objects do not overlap?

This is called occlusion. The short answer is, no right now our system doesn't do occlusion. To do that right, you have to build a high probability real time polygon mesh which we aren't building into a live product yet.

So if you want to measure up your apartment, all you basically have to do is take a Tango device for a walk around it.

There is more than one way to do this. We can do this, but it's not real time - it takes a few minutes. However it's getting faster very quickly and this use case is in the pipeline.

Dedicated sensor hardware blows away laborious fancy math to overcome the limitations of a sensor. Look at the adoption of the Kinect and LIDAR in robotics hardware. No-one is seriously using monocular vision for any kind of serious environmental sensing. Stereo vision + pattern is pretty much the minimum to get satisfactory results.

> No-one is seriously using monocular vision for any kind of serious environmental sensing

I believe Tesla uses a single camera for auto-steering.

Dedicated sensor hardware blows away laborious fancy math to overcome the limitations of a sensor.

This doesn't work at consumer scale.

No-one is seriously using monocular vision for any kind of serious environmental sensing

That's why we're different.

They demonstrated several use cases, live: furniture size checking, education about dinosaurs, distance measurements, and entertainment. Really, did you not see that?

Does Tango come with software to stitch together a 3d model with a texture? For example, could I use it to scan a shoe or other discrete object in front of a bluescreen?

That's probably the first question on any web retailers mind.

If VR/AR technology takes off (as Facebook, microsoft, and Google believe it will), this provides a convenient, accessible, and cheap means to produce content. No specific use-case needs to be established yet IMHO.

Not really. Just having a depth camera doesn't help with content creation. You need a lot of other pipelines to work.

See:

A Large Dataset of Object Scans

Sungjoon Choi , Qian-Yi Zhou, Stephen Miller, and Vladlen Koltun

Reply on news.ycombinator.com