Portal Scanning Guidelines?
AkceptorUA
✭✭✭✭
Hey, are there any guidelines for portal scanning?
Sure, one can simply record those videos and work towards a badge but I'd like to provide a significant value per each scan, not just waste the data
In-game hint asks about walking around a portal (if possible) but as I know from my previous 3D scanning experience it worth to do these things as well:
- Change distance to portal fir better texture resolution
- Capture with tilted camera
- Capture the top of the waypoint (obviously you can't capture the top and walk around)
- Capture the place of
connectionportal to the ground
In addition sometimes it's better to have two separate captures instead single one (in case of some obstacles).
In addition if the wayspot is flat (e.g. mural or graffiti) its quite impossible to walk around so we should select proper angles for starting and finishing captures.
What if some objects (that are not related to the portal) appear in the camera view? E.g. trees or just branches around the wayspot in the park... What about moving people\cars in the scene background?
Any advices of scanning wired structures like this?
Comments
Oh, I have already thought about the problems with trees and plants in the picture, also about objects that are too big to walk around them. It would be interesting to know if it is sufficient to scan single parts of a larger or partly hidden wayspot and put them together in the end or if it is really necessary to scan the wayspots as completely as possible.
So far, they've said it's best to scan portals that you can walk around (like statues, monuments, etc). But if you can't, a 180 degree scan will do. And you're allowed to do multiple scans of the same portal. They've also suggested doing scans at different times of the day, or in different weather conditions if possible.
There's at least one portal near me that's such a big building that I can't be close enough to scan it without being so close I can't get it all in the viewfinder at once or walk round it all in the time allowed.
Some guidance on what we should do for flat objects like trail markers or awkwardly shaped things like football pitches might also help us help you.
The guidelines for portal scanning can be found here: https://niantic.helpshift.com/a/ingress/?s=portal-network&f=portal-scanning&p=web
...but they don't answer any of the things we're asking here.
Yep. Those guidelines are like any programming tutorial on the Internet teaching you how to write simple
helloworldprogram. And in real life things are more complicated.Even for statues its not so easy to find one without trees and people around.
Plus we can't see the result of the data processing. I have a bunch of portals nearby I submitted scans for. But I never know if those scans were good enough, I don't know if they collected enough data points or they need more data from some specific point of view.
Basically I don't see thing like this:
So its kinda blind process
Regarding large buildings, they do say in the scanning guidelines: "If the Portal is too big or too far away to capture with a single scan, try scanning the Portal with multiple scans, keeping some good overlap between each scan."
I'm not sure what you mean by a "flat" trail marker. I've scanned several things in my local greenspace which are like trail markers. Posts with a map on one side. I've still scanned all around them, and from different distances.
I'm not really sure a football pitch is a scannable object. But isn't there some actual object that's the portal picture, such as goalposts? Otherwise the submission is just grass anyway.
I don't think every portal would be able to be made into a 3D model. For example, you could surely make a 180 degree scan of a mural, but what sort of 3D model would you want from a mural? Nevertheless, I say scan all the things. Make multiple scans from different distances and heights. Give Niantic the data and they'll give you a badge.
these are good points passed on to the team. we'll post more detailed guidelines.
And what else would be really nice is having an ability to see the scan data already present. Maybe published just by you or published by people per-portal. Maybe some hints like `this portal lacks scanning from South-East` thing to point people to the best ways of improving data quality.
And maybe some way to specify that portal is 2D structure (mural or plaque) during scanning so that processing will be simpler
@AkceptorUA yes, different kinds of structures (buildings, statues, 2d signs and plaques) call for different approaches. passed on. expect to see some clarifications.
There are two pieces of information that I'd like to know before I scan a portal (or scan it again):
Thank you for all of your feedback and input. As you mentioned, there are nuances to creating a good Portal Scan, and we really appreciate your help in communicating those to our community. Our team huddled to put together some additional recommendations or guidelines while Portal Scanning.
For those new to this feature, the opt-in Portal Scanning feature is a way to crowdsource AR maps of Portals — places of significance that are important to Agents. We blur any personal information, like faces and license plates, and then we generate accurate, dynamic, 3D digital twins of that location. With these digital twins, we can understand the geometry of that location and build a 3D map that includes all the depth and complexities of that place. In the future, this will allow our team to build new AR features that recognize and adapt to these physical, real world places. For example, it could mean the difference between hiding an AR cache or dead drop visible from a strategic vantage point versus dropping a Capsule in plain view on the Scanner map screen.
Here are some tips on how to scan different types of Portals to provide the most useful information to our team:
Additionally, while it’s better to upload Portal Scans without obstructions or people between the Agent and the Portal, we understand that this is sometimes unavoidable. We automatically check every Portal Scan to blur faces, license plates, and other identifying information.
A few things to avoid, if possible:
Lastly, these guidelines are meant to help provide more insight into how Portal Scanning works and can contribute back to our Real World Platform. We appreciate all Agents’ contributions towards our Portal Scanning initiative, and every scan is helpful toward building a more dynamic and realistic 3D map of the world for the benefit of all Agents.
Thank you so much for this, but I think you should start a new announcement thread so everybody can easily find the information.
@Atunatunak we're going to wordsmith this and add it to our Help Center article that we link to from the Portal Scanning tutorial: https://niantic.helpshift.com/a/ingress/?s=portal-network&f=portal-scanning&p=web
@NianticBrian thank you for your information. i think it is a great idea to have a 3D Scan of every Wayspot.
I have a few questions, if possible, please answer or bring them to a team-meeting
Thank you and have a great weekend
I know you're unlikely to want to reveal future plans, @NianticBrian but it would be difficult for someone not physically present at a location to produce a portal scan that matches ones previously supplied by agents. Could portal scans therefore be used as an anti-spoofing measure, perhaps requiring them for the creation or destruction of very large fields, or when movement speeds are pushing the limits of the physically possible?
I like the idea of "Due to suspicious activity, to prove you are not a bot, upload a portal scan" in theory, but the problem is that many phone models don't have the necessary hardware to do scans.
But maybe it can still interact with the results of scanned portal, and if weird interaction like walking in scanned area that is known for being impossible to walk, help indentify the cheat.
@NianticBrian Thanks for update, its definitely very clear now :)
Just a couple of clarifications regarding the camera position.
Both these approaches can help to provide scans from different angles so basically more data about the portal.
And one more clarification about the distance.
All that's really required to do a portal scan is a camera, and you'd be hard pushed to find a phone that doesn't have one of those. The official requirements are very low too. From Niantic's Helpshift page:
Portal Scanning is available on iPhone 6s and newer devices running iOS 11+.
Android: Portal Scanning is available on devices running Android 7.0+
That's barely higher than the actual requirements for running the game itself, and given that building a database sufficiently large to roll out a scanning requirement would likely take a year or two, it would be entirely reasonable for the minimum requirements for scanning to be the same as for playing by then.
As I said, all that's really required is a camera. A requirement for AR-Core probably means Niantic are offloading some of the processing to users at this early development stage but there's nothing magic that AR-Core does that couldn't be done sever side.
As for requiring internet, you're describing a (low priority) bug, not an unsolvable problem.
I suppose its definitely possible to process everything at server side (not sure though if it means send the whole video to the server - some preprocessing is still possible at client side) but my point is prefer the framework usage over creating the whole infrastructure (and pay for it, including resources consumed).
I think user-side processing is just cheaper and even better in terms of having hardware related processing (different cameras, sensors, etc) done by the framework which is aware of particular HW capabilities and sending only the unified data (kind of depth map) to the server. I think user-side processing is more or less a trend :)
One more thing I believe using client AR-framework somewhat guarantees you can't simply record fake video (as people used to do with photos for submitting portal nominations). It should be harder to fake ARCore than video camera.