OSMers, (how) do you use the Mapillary web app?

Hi everyone! I know there are lots of OSM contributors here and I’d like to learn a bit more about your workflows after upload.

I’m interested in is how often you are using Mapillary on the web and what for?

  • Do you go and look at your imagery on the big screen after it’s uploaded?
  • Do you use the editor (compass angles, deletions, etc.)?
  • Do you explore what kind of data has been extracted from it (object detections in the image, features on the map)?

Or do you rather just head straight to iD or JOSM to pull up the Mapillary layer and start editing?


For the rest, I hope you kept record of all my complaints.

Whenever there happens to be imagery in the area of interest, I use that as one of visual references.
Usually I use the big screen, but also use the plugin in OsmAnd.
Considering the general lack of coverage I usually use the editor on my own images, although could probably avoid 70% of that use if there were preprocessing options in the uploaders.
The layer in iD serves more of a shortcut, although sign detections (when they are finally fixed in iD) are very helpful. Don’t export points data into a separate file, as with limited imagery there aren’t that many points that I can’t just go through manually

Do you go and look at your imagery on the big screen after it’s uploaded?

I periodically view my recently uploaded imagery to check for any glaring geolocation errors or bad camera placement (i.e. dashboard showing too much). If I have captured a particularly interesting area I will spend more time looking through the photos.

Do you use the editor (compass angles, deletions, etc.)?

I use the editor primarily for the bearing/heading “interpolation” feature when uploading photos taken outside with an action camera. And deletions as needed.

Do you explore what kind of data has been extracted from it (object detections in the image, features on the map)?

I periodically explore what kind of data has been extracted but so far only out of curiosity. I have not incorporated detected features into my OSM editing.

Or do you rather just head straight to iD or JOSM to pull up the Mapillary layer and start editing?

I use the web app to quickly check my uploads but the vast majority of my usage takes place in the iD Mapillary layer.

Hi Katrin,

Thank you for asking. Take photos while cycling to and from ‘follow-the-arrows’ type walks, and on those walks.

1/ Use a .de route planning engine which routes over backroads and offers a Mapillary overlay so that I can drag the proposed route to take in more ways which aren’t documented yet.

2/ During the first months viewed each sequence to learn from my omissions - have now come to a point where there are many more pics, and still the occasional bit is missing, but less frequently.

3/ As more of the monthly allocated volume is needed for uploads there’s less available to download and view my pics, or use the ‘tag’ and ‘comment’ options - we’ll come to that later.

4/ no idea how to explore data extracted ; there is a ‘roadsigns recognised’ layer in the iD editor on OSM, but my mapping is survey-based with pics taken for the benefit of other mappers working on other aspects.

Some while ago stumbled on an part of Mapillary where one confirm/reject recognised signs as indeed being a particular sign - thus help Mapillary correct their recognition.

5/ It would be helpful if the Mapillary filter to include a particular contributor’s pics, or pics from a particular time frame, were available in the iD editor, as it allows to exclude -or concentrate on- pics showing the previous situation.

Let me expand on the above:

1/ Where I live all ways are covered up to at least some 8km out, and further out almost all ways have been pictured many times over the years; it would likely be helpful if that .de route planning engine had several Mapillary overlays to choose from so that one can select one of ‘ways covered in the past four years’ and ‘roads covered in the past two years’ : the regional and local governments here have been totally engrossed adjusting their roads, very much adding cycling and walking infrastructure, designated parking areas at the expense of carriageway width : this can only be amended on OpenStreetMap if there are recent pictures.

Some of those re-done roads one’ll be aware of, others are a chance discovery, but planning a route to take in roads pictured years ago there’d be a better chance of finding those roads. Compare https://www.mapillary.com/map/im/xKuLrbjstSO01hnkDs4aVQ and https://www.mapillary.com/map/im/1Ou_trzOJvk6W1to9GstuQ or https://www.mapillary.com/map/im/pkuj2Ks0CbjexBA0CHaUiA .

2/ Tend not to view the whole series anymore :

2a: blurring has become much better, all faces -and backs-of-heads- were blurred after you upgraded the algorithm, as were licence / number plates, while there were fewer erroneous blurs of railings;

2b: would need to view each pic to ensure it’s correctly placed and pointing the right way - which will take hours, but spot checks in the web-uploader before uploading shows any systemic errors;

2c: that leaves indicating which pics show items potentially of interest to mappers - by using the tag and comment modules this takes long, too long : once the series has been published need to wait while the pic downloads and is displayed, glance and decide there’s nothing to tag (as it is just a stretch of road showing ‘nothing to see here’ - adding to Mapillary’s knowledge of where there aren’t any traffic or other signs to map), or load the tag module, add a rectangle where the traffic sign or other object of interest (drinking water fountain, life ring, bin, bread vending machine is, pick from the list till one arrives at ‘traffic sign’ , download the comment module and type comment “one way, opposite direction, bicycles excepted”, then upload :

Whether it concerns 1000 or 5000 pitures : anything in that order taking as little as 8 seconds amounts to 8000 or 40.000 seconds which equals like 2¼ to 11 hours - that’s as much time as the trip took, and I for one just don’t have that amount of time , plus that it would be better spent learning the finer points of editing OSM, or expanding work related knowledge and skills through self-guided learning.

3/ Do view pics before uploading in the XnView photo manager,which on the QHD screen offers 48 thumbnails in adequate detail to spot the worst ones (which are obviously deleted, unless they hint at something to map). In that thumbnail view one can right-click on relevant pics and edit the IPTC fields - conveniently labelled Tag and Comment - to add ‘proto-tags’ like AED, bench, vending-machine and comments like “that green blob on the building on the left is/isn’t an AED - would’ve been ‘fat chance’, as it is on a primary school”. All that now remains for Mapillary is to parse the ‘tag’ and ‘comment’ lines in the IPTC section and convert AED to defibrillator.

Not clear why Mapillary prefers defibrillator over AED : the ones accessible to, and operable by, the public rather than those found in hospitals, requiring a team of medical professionals and occasionally pictured in those ‘hospital soaps’.

Also spot that - at least last time I looked - Mapillary doesn’t offer a tag to indicate tourist route markers, and local / regional / national walking / cycling / horse-riding / ATB routes and networks. There is a marker for adit, which is pretty rare in flat-as-a-pancake regions like the Netherlands and Belgium, which makes it all the more puzzling why ubiquitous route markers can’t be tagged.

4/ But that set of recently recognised road sign pics was limited, obviously a training series rather than a feed of recently recognised ‘low confidence score’ pics put in front of human eyes to confirm / reject; stopped scrutinising those until Mapillary lets me know that I’ll be confirming questionable recently recognised pics.

That’s it for now, there are other niggles, but those aren’t OSM related.
edited to add ‘t’ to ‘he’ to form ‘the’

I am aware of the difference between an AED and a ED. I don’t know if internal defibrillators exist, I hope I will never know :). But “defibrillators” is the term chosen by OSM. And for once I agree with the choice. AED is not known in french speaking regions. And you know, abbreviations change. And we cannot expect people to know acronyms. Diseases also have changing names. I also prefer AED in my changeset comments because it is shorter. In OSM it is not possible to really fully describe everything. Hence the pictures.

If there is not too much images in the area I’m editing, I tend to use the Mapillary plugin in Josm.
If there is a lot of images, I use the web app.

But, as the Josm plugin is unstable, and doesn’t support spherical images anymore, I tend to use more and more the web app.
The Josm plugin deserve more

1 Like

JOSM plugin definitely need update to support 360 images