I am looking for some constructive advice on how to successfully nominate the entrances to a local urban forest in my town (“Bosco di Via Carrer”).
The Context:
This is a well-known public urban woodland used daily by the community for jogging, walking, and cycling. It perfectly fits the Exercise and Exploration criteria.
The Problem:
Unfortunately, this park lacks official signs or nameboards at the specific access points. The entrances are only defined by wooden bollards/posts or wooden fencing that physically separate the paved road from the internal gravel paths.
The Rejections:
I have tried nominating these entrances focusing on the wooden infrastructure, but I keep receiving immediate automatic rejections (AI). I suspect the ML model sees “trees + ground” and immediately flags it as a “Natural Feature,” failing to recognize the wooden markers as a man-made entrance.
My Question:
Since there are no signs available to photograph, do you have any tips on how to frame the photo or write the description/supporting info to prove these are valid Access Points?
Has anyone had success getting “unmarked” trail/park entrances approved by highlighting the physical barriers/posts?
I have attached the photos of the candidates below.
The photos you have included seem good for the supplemental.
I would also make sure to tell reviewers in supplemental to look for the marked trail on Google maps. The submission may not get to human reviewers, but appeal reviewers can read that statement.
Can I ask you one last thing? What do you think about these? They are old stone markers from the local aqueduct connected to the land reclamation as here everything was a swamp.
I can’t make a judgment just from photos. I wpuld need more explanation about how these meet the criteria. I’d also want really good proof of what they are.
For example, the web submit allows you to upload multiple supplemental photos. Maybe one of those could be a rubbing of the stone to show what the writing says since it is too worn to read by sight now. And then google this project or these marker to get more information about them.
I like finding old boundary markers. It can sometimes be a challenge finding what they were for, but the more information you have the better. I’m surprised they were rejected by the ML system.
Too generic, consideration given if there’s a trail marker &/or titled signs. But trails are tricky as many can’t be viewed for confirmation via Google maps feature, as such I’d check i dont know for accuracy with location as the reason. Best
This is an example where using the Web Submission process will help, as you can add up to 5 supporting images. For trail markers not visible on streetview, use these photos to prove location, e.g., a series of photos leading to something that is visible.
First check that the multiple-photos feature is working, as it has at times (possibly at the moment) not worked.
If the worry is getting past the AI do we know if adding more pictures versus having just one is safer? For example, would the AI reject if there were four pictures it liked and one it didn’t even if humans would have understood the context of what the one photo was showing?
My strong expectation is that the ML system uses the main photo when deciding whether to reject a submission, but not the secondary photo.
Since the secondary photo can have people’s faces and license plates in it, which are both extremely ill-advised for the main photo, and doesn’t have to include the POI in it (even though that is definitely best practice) or be of anything in particular, I severely doubt the secondary photo can be used in any way by the ML system except possibly to confirm location (but even that I have trouble imagining that a failure of the secondary photo specifically to prove location could result in a rejection by the ML system).
The secondary photo can even be a mashup of multiple photos joined together, which is what some people did to prove location before the multiple-photos was available through web submission. Reliably parsing that through ML would be a headache.
If the secondary photo is used to confirm location, then it would be extremely illogical to insist that every secondary photo must all individually confirm the location, so multiple photos could not cause any harm.
Therefore I would have no concerns about the ML system with having multiple supporting photos.
