Firstly, thank you for your human answer, it is appreciated much more 
Then that one is not a problem, I was just not sure if it was really taken on the foot. 
Professional cameras (or regular smartphones with apps like Baba or regular camera apps like OpenCamera making timelapse photos) will tag each picture with different location (assuming you were moving).
Your app however seems to tag them all with exactly the same location, which is problematic, as user can’t know where is each picture taken (which picture was before the crossing, and which after? In which direction is the user going, i.e. is the next picture more to the North, or more to the South? Or have user turned to the West? etc).
And if the user is not sure which road is corresponding to which picture, then it is useless for them for OSM mapping purposes (e.g. in your example, which road is that gravel one, and which is 2-lane asphalt road? To which direction is there forbidden traffic? They can’t all be “the one directly to the North”, and yet that is what they are claiming)
1 or 2° mistake would not be a problem. But here problem is much worse.
I.e. all your pictures in panoramax are claiming to be pointing directly to the North (i.e. 0° Azimuth). And it seems incorrect, as direction of e.g. this and this and this and this pictures seems wildly different. Like, 70-290° different; not 1-2° different.
And that is the big problem as mapper trying to use Panoramax cannot say what they are looking at. And without being able to know what they are looking it, the content of the picture is close to useless. Especially when combined with the fact that we don’t even know the position of the user either (are then in that picture North of the crossing? Or South? Or West? Or East?)
It gives the user some orientation about how far the pictures are far apart (which is especially critical as they miss location of each picture). If we know user was walking (and we can guess in which direction), and we know how many seconds is between two pictures, we can estimate how far the user have moved and where the next picture might actually be.
And big part is that you actually can do something about timestamps at least.
If you know video was 81 seconds long, and you know it started at 12:49:00, and you extracted 27 pictures from it (e.g. one each 3 seconds), then you know if the first picture is at 12:49:00, the second is 12:49:03, the 3rd at 12:49:06 etc.
So you can record that different timestamp information in EXIF of each .jpg picture. It is not absolute time precision that is important (i.e. it is pretty much irrelevant whether the picture was taken at 12:49:00 or 12:53:00), but what is important is:
- relative time between pictures (e.g.
00:00:03) as it indicates (esp. in combination with mode of transport) how much meters is likely between two pictures; and
- general approximate time (as that, in combination with general location, with some detective work can help determine in which direction we are looking – e.g. by knowing in which direction the sun and/or shadows are being cast in that country at that time)
Uhh, mine does (if we’re taking about “EXIF tags”, and not “Panoramax sequence tags”). As does any regular smartphones with pictures taken with Baba app mention above, or OpenCamera, or other.
See e.g. this 1st picture and 2nd one. They are taken at same crossing, but show different roads, and their direction is very important – otherwise one could map “forbidden right turn” on the wrong road!
As you can see in their details, there is GPSImgDirection EXIF tag in each of those .jpg pictures which defines it.
That is what I was trying to say. Without correct location and direction and timestamps, they are (almost) completely uninteresting for OSM mappers; as they give mapper more confusion then help, especially in your example when there is a crossing, and so it is unclear which picture represents which road.
If that was a video of a single road, with no crossing and no changes (i.e. all pictures basically the same), then there might’ve been some little utility as:
- there would not have been any confusion about which road we are looking at (as there would be only one) or where some feature was (as there would be no additional features), and thus
- user could extract some information like
surface=asphalt and lit=no and lanes=2 from it (and maybe even smoothness=good if quality was good enough – which in this particular example it wasn’t).
But as majority (if not all) of that information can also be seen in many regular aerial imagery (which already covers the whole world), the pictures put on Panoramax would not really produce much (if any) additional value.
Thus I suggested adding at least tags like guessed_locations=yes to your Panoramax sequences, so users can know what problems to expect from them, and not trust them as authoritative source of precise information (as with telling tags they’d known their information is ambiguous/approximate/interpolated, so they would know to only use it as a supplemental source).
Hopefully I managed to clarify the issues (that’s why the post is more on the verbose side, hopefully enough to make things clear). Let me know if there is still some confusion remaining.
).
icon which will translate it to my preferred language, just as it will translate my post to your preferred language.
