Let’s walk through the list of things I’m told “wayfarer criterion” might mean.
To begin with, it’s “our team” and rejected within 24hrs so we can assume that it was rejected by automation.
First possible meaning, failing to demonstrate social, exercise, or exploration criterion.
This submission is of a playground. I’m taken to understand that playgrounds have already demonstrated at least two of these and sometimes three.
I’ll also point out that the playground next to it, submitted within minutes of the other, also shown below, has apparently passed automated filtering.
Second possible meaning, photo with a lack of central object. I’ve been making a clear point of including these in every submission. I don’t think this is lacking here. And again, the adjacent playground is substantially similar and was not rejected.
So… what else might “wayfarer criterion” mean?
Rejected submission:
Similar but unrejected submission:
Putting on my computer brain…
Those are photos of a fence with a sign. Those are not photos “of” playgrounds. In Wayfarer terms, they “represent” playgrounds. A human brain can understand that. But i don’t think ML should know that.
I’m also uncertain about whether it reads your text and sees “elementary school” and says “better safe than sorry” and denies it. I try to avoid discussing words that could be a straight up rejection reason in my text (not sure that matters, though).
The one that ML allowed through is better framed.
4 Likes
I would also avoid using the word “School” anywhere as the computer brain will flag that and think its part of a school which breaks guidlines.
I would say its a playgroud attached to a libuary. (which it is) but leave out the bit about it being a former school playground.
Er… you don’t think it’ll be rejected for being a school in that case?
I think the computer AI model is mistaking the POI as a school and then rejecting it acordenlly. As youve said it was rejected within 24 hours which sounds like an AI reviewed it.
I would mark it to be reviewed by Niantic Staff as that gets a pair of human eyes on the submition and they will then at least provide more detailed feedback or authorise the POI
My first thought is that ML sees a picture of a fence. But it has a sign, and often signs are the wayspot.
ML can read the words on the sign. Maybe it doesn’t like “ages 4 to 6”, or “supervision”.
Also, the colors seem to be very enhanced (so the fence lines are jaggedy), so maybe ML doesn’t like that?
In reference to the adjacent playground that was approved, could it be that ML sees this second playground as a duplicate, and rejects it?
They have different locations. And I’m pretty sure they’re far enough apart to be distinct.
Except… the second one wasn’t rejected and has nearly identical wording.
I wouldn’t mention this being a former school, and I would try to take a photo of the actual playgrounds, not the signs. I nominated, and had accepted, 2 playgrounds for different age groups at a park. I also gave them titles that didn’t include the ages, as those can be included in the description, and included the park name; the library name would be nice to include in the titles.
2 Likes
I would like to point out and highlight that this is yet another situation in which the phrase “wayfarer criterion” communicated absolutely nothing.
I mean, I appreciate the pointers that might make this acceptable on resubmit. Those are helpful. But it seems pretty clear that no one, not me, not the other helpful folks, has any idea what “wayfarer criterion” means in this case.
I’m beginning to suspect that it is Niantic speak for something I probably shouldn’t spell out too explicitly here.
Working as designed.
First of all, AI doesn’t give detailed reasons for things. It’s more holistic than checking IF/THEN/ELSE hard specifications. Have you read about AI diagnosing things on medical images - then the human experts have to backtrack to see why AI could flag it sooner?
Second of all, even if AI would give a specific reason - Niantic doesn’t want to give away any more of how AI works than they have to. Bad actors have already figured out how to “game the system” - let’s make it harder, not easier, to mess up our databases.
Lastly, they do want us to brush up on the criterion - to read, ask others, brainstorm.
1 Like
That’s patently untrue. AI can absolutely give detailed lines for it’s reasoning. That’s a huge part of it’s purpose.
We use the ability to backtrack as a primary criterion for selecting models, in fact.
ML certainly helped me brush up and refine how I submit nominations and edits. Learning from the community of what ML likes or dislikes (example is that ML sometimes dislikes photos with too much greenery or even snow) and how to improve on these aspects is helpful.
Oh! If some folks are thinking that “photo issues” are not “wayfarer criteria,” then please check the criteria tab again: Niantic Wayfarer
It specifically lists under " Photo Guidelines":
- Well composed - the Wayspot or placemarker is centered without too much foreground or background, or objects passing by in front
1 Like
Yeah, I see where Niantic would say, that a picture that looks all grass-and-sky doesn’t make a great Wayspot. When the object is such a small part of the picture, you’d never see it on your phone in game - or on a map pin (optimistic that we’ll get a combined map someday).
ML is teaching people to focus on the thing. That’s good!
Sure, picture can be updated, but it’s much more cumbersome to do (so less likely to happen) than, say, a text edit. We’ve all seen pictures still primary from 2014 with car hood and rain on the windows. If ever fixed, it could be years. Someone has to be there on a sunny day, get out of their car, make their spouse wait out of the way, worry about bystanders, etc.
2 Likes