Refreshing existing coverage

Hi all, just wanted to ask if there are any guidelines or ‘rules’ for contributing in the areas, already covered?

  1. Does it make sense to upload imagery in the areas which were already covered?
    After reading the forum - I assume the answer is yes, but I am still looking for some guidelines.

  2. For the areas already covered - I think it will not make any sense to upload new sequences every day or week, or even month. So what would be practical time to allow before uploading new sequences? 6,12,24 months since the latest sequence available?

  3. Imagine I have covered the same area with a better quality (better weather conditions, better camera, etc.). Should I prefer keeping my older sequences intact (i.e. for historical records) or should I better delete and replace old sequence/(s)?

Anyway, what I would like to see in the tool - is a user-friendly interface to filter and delete own sequences.
E.g. ability to select all my sequences in the given time/date interval and an option to bulk-delete them.
This will help cleaning up a lot.

1 Like
  1. I haven’t seen official guidelines, however the fact that there is no moderation for unfocused/night time images and functions like Time travel makes me think repeat capture is encouraged. It’s more data after all
  2. Sometimes even a day is suitable - temporary road works or poor weather, and the next sequence is completely different then. Play it by ear
  3. Images from older sequences are useful for time travel, and also as a gps track of sorts, the more the better. I’d keep them.
    After all, mapillary might build filter functionality, but restoring deleted photos is a more challenging process
  1. Data is king and more data… is more king :wink:
  2. if it’s the same there isn’t much value. But a different time of the year (Winter will give much different results than summer) or when a new house is (being) built in the street or other construction (or demolition) work… or better quality images… if it’s different Mapillary will be happy to see them!
  3. if what you have taken images of is exactly the same and only the quality of the imagery is better… i’dd vote delete. But if in a broad way at viewing at it there could be value to the old sequence… let it be.

I am working on a tool for me to better catalog and filter my own sequences… when I have something useful I’ll post about it in this forum. Can’t help you with deletions… but the web interface of the app of Mapillary is quite powerful!

1 Like

When I see changes like road works, building construction or whatever, I take new sequences.

Also, last year, in France, they changed driving maxspeed in rural areas. But the devil is in the details and some portions are differently affected. So, I’m tacking new sequences of all important roads that don’t have sequences newer than July, 1st 2018.
Before :
After :

I also take picture when snow is on the ground:

I would also like to record sequences at night to map street lighting but quality is too poor:


Nice (I’dd skip the nightly recordings though… the street lighting is also visible during the day :wink: )

My two-bits

Do It

Even crappy pictures for the algos are rich in data they can harvest.

I have personal interest in astronomy and light pollution is an issue. Modelling light pollution is an interesting topic that need precise description of light sources (OSM would be the good place for such description storage). Night pictures would give much more data about light type, directions and so on… Just a dream :wink:

1 Like

Making a contribution requires some effort as you could imagine.

I am really keen to hear a comment from one of Mapillary team members.

Just want to be sure there is a REAL use of multiple contributions HERE AND NOW for the project, not something theoretical (time travel, etc)
I doubt there are any algorithms analyzing minor changes (e.g. new sign post or similar) and I doubt community members will be interested in analyzing 20 sequences in the same spot.
Moreover, I personally find it very complicated to find the most recent capture in the high dense areas (i have to look through each of them to find the most recent recording and personally and practically I am only interested in the most recent one since I don’t care what was there 1-6-12+ months ago)

My humble opinion.

1 Like

@gpsmapper, I’m not sure what you’re getting at. It sounds like you want some hard and fast rules. There are none.

Maybe you could put something together based on feedback from others?

Don’t get me wrong, I am just trying to find any evidences that this (recurrent coverage of the same area) is being used in real life by anyone. So far I’ve heard only ‘more the better’ without any real case confirmation if and how this data is being used.

In other words - is it more ‘nice to have’ or more ‘must have’

I will ping mapillary team by mail, so hopefully they will share their thoughts on the subject.

IMHO - The most important audience for these pictures are the algos, not humans.

1 Like

When you spot new traffic signs, or a changed road layout then recording that is obviously useful.

Re-documenting a road every so often when nothing has changed seems a waste of effort, but can serve to confirm that nothing has changed - in which case you might upload just one in four or even fewer: it’ll show ‘nothing to see here’.