A week ago it finally processed into my sequences. Today I find that no Traffic Signs and Point Features were detected. Has the extraction functionality been disabled, is it taking longer, or what is actually going on with it?
1 Like
Can you share a link where you are seeing this? cc: @adalvit
I forgot to add that I mean objects detected and added to OSM. You can preview them in Rapid/iD, in edit mode, turn on the Mapillary layer, and by selecting the objects there, you can see which photo it comes from. It’s similar in JOSM. There are no detected objects on any of the streets I drove through that never had a Mapillary while recording the sequence. e.g. “Stanisława Augusta” street in Warsaw (Poland)
I found out that objects appear 3-4 weeks after processing is completed. I can confirm that I have started noticing these objects in some of my sequences.
1 Like
How to report incorrectly detected road signs? Different sign on the map than in the photo?
No easy way, you can try emailing support.
I also found object detection quite slow over the last year or so
Traffic sign detection: I mapped two roads 3 weeks ago. One with from a bike, one from a car. The traffic signs are detected in the pictures in mapillary. But only some are detected in map view, available for download.
Any suggestions? I see other people have waited a long time to get the traffic signs. My main goal by contributing to Mapillary is the option to detect and download traffic signs.
Here is a link to an area I photographed in the middle of march. No traffic signs detected here:
There are older pictures on this segments, but the traffic signs have changed since.
Regards
Sveinung Bertnes RĂĄheim
Bodø, Norway
@tao - are you able to chime in on this one?
Nothing happened on this one. There has been no more traffic sign detection on the uploads. I did some new mapping 3 days ago. So far, so signs are detected for download. Most of the signs are detected in the pictures while running through the pictures.
For me, traffic sign detection is the main purpose with mapillary. And when this is so unstable it is hard to consider using it on a larger scale.
Any update on this @boris or @tao? We are about to launch a fairly large scale mapping solely based on the purpose of collecting traffic signs, so for us it is very important that the detection algorithm is working and that it won’t take 2 months for the traffic signs to show up in the platform for downloading or the API.
For example, this image from 2024-03-16 detected a new 30km traffic sign shown in the image, but it doesn’t show up in the map for download. This is more than 2 months. Could you also confirm if it doesn’t show up in the map that it wouldn’t in the API either?
thanks,
Diego
2 Likes
Dear @tao @boris @abalys and other related mapillary heroes…
Any news on this topic ?
We are also about to flip new collection of traffic signs in our system.
/ Nash
1 Like
Also cc: @abalys for ideas
2 Likes
I have noticed similar behavior in that it takes considerably longer for traffic signs to pop up on the Mapillary layer than since about a year ago (maybe longer). Traffic sign detection itself is quite quick and happens right about at the same time as blurring does.
However, for me far more important is the triangulated positioning of traffic signs on the Mapillary layer and that their positioning precision improves with additional sequences over time. Same goes for traffic sign consolidation, although I am aware that this is no trivial problem to solve with imagery captured at different points in time. I just wish that we would get back some sort of traffic sign consolidation because currently some intersections look like just a giant blob of colorful mess. Anyway, thank you guys for your work.
1 Like