No offense but this is a really poor argument because one image (even a 360° pano) for more than every 15 meters is poor quality either in an urban environment. So please, hold your horses there.
Since currently only UKMs are scored, there is a loop hole in the contest rules where one can easily eat many UKMs by capturing long sequences with just one image per intersection or junction and move on. Hence, I do call all participants to do exactly that because according to the rules it is a fair strategy to advance your chances of winning. Seriously, go for it people!
Like I have mentioned before, the scores need to be leveled by a target ipkm factor. mapillary_tools’ 0.333 ipm (currently for video) seems like a reasonable and fair target to me. A target ipm factor not only ensures a level playing field for everybody but also that participants do not overshoot or undershoot.
For even more fairness, you can add total pixels and pixels per meter factors too.
Thanks, everyone, for the strong third week as part of the CompleteTheMap Europe challenge. I’d like to share the UKM stats as of July 7thth, 2025. The challenge has helped us acquire 2593.19 fresh UKMs for improving maps, with a total of 258,444 imagery!
On the email you’ve sent at the begining of the challenge stated swag would be elligible for those who have 100-150km and 150+ ukm would qualify to a amazon gift card.
Meanwhile the blog post states that swag is only available for those who make 200-300 ukm and amazon gift cards above 300 ukm
With all due respect but that’s the total opposite metric that is supposed to be used here.
The goal is not having an image every meter. It’s supposed to have as much of a city with street imagery.
I’ve noticed that you’re not even in the challenge. Not only that. your profile is filled with 100~200 m sequences with thousands of images. Some of them have an image every 20 cm. Most of them don’t add value to the others and won’t be more useful in the context for mapping openstreetmap and HERE maps which is one of the goals of this challenge
Let me clarify your undifferentiated confusion. My comments serve as an inspiration to think about fairness and a loop hole in the contest rules. Nothing more, nothing less.
No, it is not the opposite. It is a proposition of refinement for fairness and quality. The ipm target factor does neither invalidate nor remove UKMs as the final metric on which participants compete. The ipm target factor levels the playing field. Why? Because some contributors capture on foot, some by bike, some by car, others by scooters and yet others on wheelchairs with different kinds of cameras. It is only natural that each mode of transportation results in different capture densities and different distances traveled in a given period of time. Hence, to level the playing field for everybody it would be wise to incorporate the ipm target factor. Mapillary has decided on the simplest metric to measure (and understand). That’s okay but this simplicity is unfortunately prone to lack full fairness for all participants. Keep in mind that a target factor penalizes undershooting as well as overshooting. Imho one image every three meters is a bit too sparse in an urban environment. Personally, I would go for one image every meter (1 ipm) in the city. One image every three meters should be good enough in rural areas and on highways. But, that’s just my opinion. Mapillary has settled long time ago on one every three meters mainly because it produces nice enough transitions in the viewer and provides acceptable coverage density, which is a fair consideration.
Right, it isn’t, and neither is it with a target density factor of 0.333 ipm.
Right, which unfortunately none of the paticipants have actually met, except for maybe @southglos who so far comes closest with 0.25 ipm.
So what? Is it forbidden to think about fairness and quality?
Actually, some sequences have a space density of 4 cm, which granted, is probably overkill (espacially given the MAX’s poor image resolution in this capture mode) but these should be rather treated as tests.
Mapillary’s dataset does not serve OpenStreetMap or HERE Maps mapping exclusively. The dataset also serves as training data for partially autonomous (level 4+) vehicles (a cheap substitute for immensely expensive LIDAR data), for measuring real estate lots, documenting construction site progress, identifying and proving real estate tax claims, and many things more. Every use case needs different levels of capturing density in space and resolution. In other words, it does not hurt to go for high density and high resolution. Anyhow, for the sake of the contest and quality of the resulting data, it would be nice and beneficial to everybody if Mapillary would incorporate a target density factor into the rules.
