I’d just like to be upload files again.
How long before it’s fixed?
My main reason to use the web uploader is I can check and correct my image gps positions before upload (Pc linux, no desktop app available)
Use the web uploader because, at least in the past, I could just upload my images and easily work with them as needed.
The desktop uploadr, at least windows current version, is still a mysterious black box. I stopped using it when I realized pictures and sequences were missing. Not suprising, like this right now, 1/3 of the images to process and uploaded won’t get pulled in. Even though it only has 18 errors? What about those others? It doesn’t say. And all those others have the proper exif, etc.
2 Likes
I now also find out that my uploads (which I limit to 999 files, already filtered on duplicates) are automatically splitted in multiple sequences when the server detects a time gap between consecutive files.
This means that when I have to wait 2 minutes at a traffic light, my upload is already splitted. Not very happy with that feature, I like to see my complete route instead of multiple small snippets of routes
I tried today a batch upload of 3500 images using web uploader, they all came through. Although the images were cut automatically into several sequences (one had a 1000 image size), no images were lost this time! Thanks, yhis means I dont have to limit my upload batch size.
But please adjust your sequence-cutter ‘time gap checker’ so it will not directly cut my trip when I wait 1 or 2 minutes or so at a traffic light. Put your threshold at 5 minutes or something 
1 Like
Was there ever a comprehensive description of this new system?
Sequences appear to be cut more aggressively and on upload, in contrast to the previous sequence bot approach. It also appears to be using time between images (with a fairly low threshold) to cut them up.
1000 seems to be the new max sequence length.
On the plus side: sequences from gopro seem to be interpolated (direction normalised) on upload or soon after - finally! I still normalise it just in case though, since when opening a sequence in Edit mode the old (non-interpolated) directions (0 deg) are still shown. It also does not seem to work if there are no other images in the vicinity to calibrate SfM
2 Likes
I have to say I am liking the new aggressive cuts less now that I have to normalise+offset more sequences manually, especially since the upload sidebar resets every time.
1 Like
Same happened to me, about 3000 images were uploaded recently but never appeared even in my uploads folder. That’s so frustrating that I even don’t understand what’s the point for all this hassle any more.
I am not a consumer of this kind of information, but rather enjoying watching how the map is being saturated with data captured from my uploads, while Mapillary definitely benefits from them in a much more monetary way (ahem, Facebook). Still, it seems that the developers use every occasion to frustrate contributors and make the whole process worth and worth.
Desktop or command line uploaders never worked for me, I had to pre-process my images with the command-line tool and then upload via Web Interface. Now it is so inconsistent that I would rather stop capturing new sequences and uploading them for good, rather than go through this headache again.
1 Like
I must, because mapillary_tools is python 2 only, and ubuntu has chosen to completely remove python 2 from its package repositories, which makes it an extreme hassle to install python 2 Support for python3 · Issue #374 · mapillary/mapillary_tools · GitHub
Would be great if one could tick a ‘normalize sequence’ box, both in the web- and in the desktop uploader.
Do I ever look at my own pics? Well, have a contract with limited bandwidth, most of which is consumed just by uploading, thus not really capacity to download to view my own pics.
Go over each journey’s crop of pics after copying to my computer : first to weed out badly exposed or framed pics; second to weed out any apparent traffic violations; third to place notes on OpenStreetMap of things to amend - either myself or with the help of a more experienced mapper; fourth to omit any long stretches which would only say ‘see, nothing to see here’.
Having done that there’s no apparent need for the time consuming normalising of sequences, especially not if the total crop of say 9k pics has been chopped into smaller chunks - some just happen to be like half a dozen pics because only need to show one lonely traffic sign.
Suggestion : show the full batch as one batch rather than as dozen of smaller sequences to the logged in contributor for normalising, please - would that really add to the server load compared to specifying each separate sequence later on?