Mapillary Vector Layer showing up signage location and not image location

[oini]

Ref #53 and #54

Mapillary seems to have updated it's traffic-sign vector layer with the probable location of the object (traffic sign) rather than the location from where the image was taken. This is great except that without the knowledge as to which side the sign is facing, it is quite difficult to gauge which two roads the restriction is applying to. Also, many a times, this detection is not accurate. Thus, it would be of extreme help if we could see both the location of the object as well as the location from where the image containing that object was taken in the osm-navigation map itself. This would simplify and fasten up our process of adding turn restrictions manifold.

• Example of an Incorrect Detection: http://mapbox.github.io/osm-navigation-map/#17.72/49.25876/-123.02055

• Example of a Correct Detection: http://mapbox.github.io/osm-navigation-map/#19.14/45.41285/-75.71271

In both the above cases, we had to look around in iD and JOSM to get the location of the same image containing the traffic sign to try and add it to OSM.

cc @planemad @gyllen @maning @geohacker @srividyacb @abhisheksaikia

[gyllen]

So we are in fact showing the real object location now and not where the user stood when they took a picture of the object.

A simple way to solve direction would be to drawlines from the object to the images where the objects where detected.

Actually above location seems pretty good, you can see that the object is located on the other side of the crossing from where the image was taken.

If object is only detected in one image then the object will have the same location as the image itself.

[planemad]

Tracking on the feature ticket