google-ar / arcore-android-sdk

ARCore SDK for Android Studio
https://developers.google.com/ar
Other
4.96k stars 1.22k forks source link

Request for ARCore roadmap #27

Closed alexeypodolian closed 6 years ago

alexeypodolian commented 7 years ago

Will you create some kind of roadmap for ARCore ?

nathanmartz commented 7 years ago

What kind of information would be most useful to you in a roadmap?

alexeypodolian commented 7 years ago

I guess that it would be cool to have a few section.

A list of features that you are working on and a list of features that you are going to add in a near future.

Also it would be cool to have some information about release dates (e.i. when developers will be able to use ARCore in production applications).

Thanks in advance :)

freezeepopprice commented 7 years ago

I've tried things out on a pixel device. Would love to know when new devices are being added. LG G6 is my development device

dylanbaxter commented 6 years ago

+1 for G6 support. Take pity on developers that couldn't afford a Pixel or Galaxy 😀

d-tan commented 6 years ago

I'm hoping that they also support the S8 plus. I found out shortly after I bought it that it was not supported :( . Really would like to know what devices will be supported next.

IARI commented 6 years ago

I would love to see a roadmap.

Is anything like Marker-Tracking (ArToolkit) or even model tracking (VisionLib) planned?

grozensmith commented 6 years ago

Markers and objects are definitely important in AR! I would be interested to hear what types of use cases you're imagining for either the marker tracking or the model tracking components. Anything in particular you would want to use those features for? Anything you like/dislike about similar APIs?

IARI commented 6 years ago

Use Case: E-Learning with AR I want to be able to develop a AR application that teaches information about systems which are comprised of multiple components each of which have their own location within the system. For example:

For each of these scenarios, in order to associate system components with their respective location, it would be desirable to be able to use marker- and/or model tracking. Model tracking might be suitable, if it is feasible to obtain suitable models (i.e. of the complete system, if its components are stationary (i.e. car) or single components respectively). Otherwise it is good to be able to support placing and recognizing markers (i.e. aruco markers).

Whenever a single component is recognized, the application may show an icon that allows the user to obtain information about the component (purpose/usage etc.).

Problems with existing API's:

grozensmith commented 6 years ago

@IARI This all makes sense. Is it clear how you would use something like markers in concert with 3D objects for which you don't have 3D models? How would a marker really work with a car or coffee machine?

P.S. Sorry to take so long to respond. I read it a while ago, but didn't write back until today!

IARI commented 6 years ago

Probably for a car or coffee machine you would prefer to have a model. However I can still imagine a particular model of some product which is only exhibited for training purposes on which you would just put markers. In a factory setting, where markers could be pinned on multiple locations, that would be more interesting.

inio commented 6 years ago

(replying to the original question)

With the release of Developer Preview 2, we've gotten much closer to what we hope to ship as ARCore 1.0. We're hard at work to support 100M devices with the launch of ARCore (coming out of preview) in the coming months.

We can't promise any specific features for 1.0 or beyond at this time, but we're hard at work on exciting stuff, so keep your eyes out for announcements throughout 2018.

chucknology commented 6 years ago

A schedule with milestones would be a good roadmap.

fredsa commented 6 years ago

An up to date list of ARCore Supported Devices can be found here: developers.google.com/ar/discover#supported_devices

sambuev commented 6 years ago

Where is a roadmap?

haroldmthyng commented 6 years ago

I know this has been asked and requested numerous times, but are there any updates regarding timeline on expansion of the 24 hour limit for Cloud Anchors?

inio commented 6 years ago

@haroldmthyng We have two open FRs in this area. Subscribe to them for updates when there are any announced.

94: general relocalization to a previously built map (akin to Tango ADFs)

523: Extend time limit on cloud anchors

fredsa commented 6 years ago

The best place to regularly checkin for ARCore updates is currently on the Google AR and VR blog.

If you prefer messages pushed to your inbox, there's also a Google VR/AR Developer Newsletter.

As mentioned above, an up to date list of supported devices is maintained here: https://developers.google.com/ar/discover#supported_devices

richardrevesz commented 3 years ago

It would be very nice to see if (and when) object tracking, body tacking and occlusion is on the roadmap. Also, the work in progress developments to see what to expect. Unity integration would also be very informative, but I assume that's up to Unity to provide.

IARI commented 3 years ago

The company i worked for more or less dropped the AR topic in 2018/2109 - I now work at a startup where we heavily rely on Facetracking. So again, i turn to this issue, to realize: It was closed, what we have is a link to the page https://www.blog.google/products/google-ar-vr/ ...

When I search the article headlines posted there within the last 3 years with ctrl+f None of the following words have a single match:

A blog like this is not anything like a roadmap - a roadmap is about planned features for the API. I claim that someone who comes looking for a roadmap will find nothing of good use there.

I don't want to come across like i just want to complain, but I have to say, the general situation disappoints me a bit. The whole face tracking stuff is bad enough when compared to apple and their TrueDepth Camera... But why can I not even find a demo app for ARCore Augmented Faces on the app store? Am I supposed to assume that it simply is too bad to stand any comparisons with something like OpenCV?