I’m using a RTK GNSS receiver (SparkFun Express) connected to an Android smartphone. I capture pictures using Mapillary app. Back, I’m uploading pictures using the app itself and wifi. Time to time, the online processing fail on sone sequences. I’m reporting each failed sequence using email to zendesk. Indeed, I don’t get appropriate answer in that I’m still loosing sequences
Hi Eric,
This particular sequence failed in the initial processing stage. Similar to one of your previous failed uploads, because this sequence failed more than 7 days ago, we are unable to retrieve the data for it, apologies.
So, usually, they answer too late.
Once, they answer quickly :
This sequence (Cluster ID: …) is showing an “invalid GPS data” error. Make sure that your device’s GPS settings are calibrated properly prior to capturing your sequences.
So, I asked for more details:
Can you be more precise on “invalid GPS data”? Does it concern the whole sequence? One or few pictures?
[…]
I’m still waiting for answer… and, of course, it was 7 days ago, so they will not be able to investigate.
By strange coincidence, many different people had problems with their sequences over the same period of time
I take photos in OpenCamera for it, and I have it marked that it only takes and saves photos with GPS location. For this I copy them to my computer and send them via Desktop Uploader (which showed all the photo and route without problems).
I measured one sequence, sent it a second time after a month and it was “Failed” second time.
One point for me is that I have not much time a home to transfer pictures. So, back home, connecting to local wifi and starting the transfer from the smartphone is very convenient. In opposite, copying GBs of pictures to desktop PC (usually after extracting SD card from the smartphone because USB transfer is too slow) and later uploading sequences one by one with desktop uploader is a waste of time. It is even worth with OpenCamera because with my smartphone, coordinate are truncated in OpenCamera so I do have to perform geotagging again on desktop with GNSS track, using josm. And, there are time shifts so, for each sequence, I have to adjust time bias using comparison with orthophoto.
Guess I’ll leave this here… the second time my gopro messes up the calendar date in the images, and I notice when it’s posted. Is it possible to somehow edit the exif on the site?
why some of my uploaded sequences (uploaded with command line) are failed, and some are not? What will happen with the ‘failed’ sequences because I can see them on the map ? Pasteboard - Uploaded Image
@nikola Failed sequences are not reported any longer. Before, failed sequences quickly became red in the feed. Now, it just says Processing 2 captures from Dec 5 to Dec 10 and the only way for contributors to know that something went “off the rails” is to wait some time until they can assume that a sequence is stuck because it does not show up in the feed as processing. This is especially annoying if you want to re‑upload a failed sequence but you have to wait some ambiguous amount of time to consider a sequence failed before you can re‑upload for sure. It is just a waste of time and very inefficient for any contributor’s workflow. It would be nice to have some sane feedback here.
Failed sequences stay in the feed for two weeks but, if we have some delays in processing, the time window to check the failure might be shorter, which is not ideal. I’ll talk with the team about making processing failures more clear and actionable.
@czecko Can you share a link to one of your sequences in this state? It would help us with debugging and improving error reporting. Thanks!
Wait, what? So I could have missed failed sequences and never noticed? I just checked recent uploads and there are 6 failed ones and there is no notification about this. And I thought retries were now automated and better internally reported?
Yes, this was quite annoying before. But now there isn’t even an indicator. And this is made more annoying because sequence browsing UI is so incredibly bad and you have to scroll endlessly with no easy way to spot gaps, search, filter, tag or organize anything.
So, one of my failed sequences from Processing 2 captures from Dec 5 to Dec 10 suddenly came back from the dead yesterday and showed up in the feed, which was nice but already too late because I have uploaded it again. Then, I had to resolve which duplicate to keep and which to delete without having an upload timestamp at hand. Now, I am down to Processing 1 capture from Dec 5, so I am hoping for the best.
It would be nice to be able to manage failed uploads at the top of the feed, like being able to delete it and have some sort of massage why it failed (like “Broken Exif”, “Cluster down”, or something). In other words, uploads should be fed to the feed by upload timestamp initially. When an upload is deemed okay it should be fed by capture timestamp as a sequence. If sequence processing fails for some reason after that the sequence should be fed by upload timestamp again with an error message. Otherwise, this convoluted Processing X captures from XXX message is not very helpful.
@GITNE , when you see a message like “Processing 2 captures” - that means the process is in progress, working as expected. So the fact that a capture from Dec 10th showed up on Dec 11th is expected (I would not call it “came back from the dead”). In general I probably wouldn’t recommend re-uploading unless more than 2 weeks has passed since you uploaded and the sequence still isn’t visible. Instances of this should be quite rare these days, and we still have more work to do to optimize here.
“Processing 2 captures” - that means the process is in progress, working as expected
Thank you for the clarification, good to know.
“came back from the dead”
What I meant by this is that sequences which I have uploaded after this stuck sequence on Dec 10th got into the feed before it finally showed up too. I am aware that processing does not happen chronologically but based on the messages and the feedback we contributors get from the system it is difficult to assess if a sequence has failed or not. As I mentioned above, the feed used to display failed sequences in red. This is gone. I continue to have a sequence in processing since Dec 5th and I strongly suspect that it has failed. Well, I am pretty sure it has failed by now. Yet, the feed does not display it as failed and I have no way of deleting it or doing anything else to it. And, judging by your statement, this sequence from Dec 5th will eventually show up? I am not so sure… but I may be wrong.
I am also aware that it is a challenge to reliably run a service at this scale and I am happy that it works most of the time. However, the very few times when something goes wrong it is sometimes difficult for contributors to make an informed decision.
This is good feedback, thank you. I believe things are supposed to work like this (@abalys keep me honest)
When you upload things you see the “Processing x captures” message
When things have finished processing they show up in your feed and the processing message disappears
If something fails processing it should show up in your feed as “Failed” and the processing message disappears. Failed captures will be displayed in your feed for 2 weeks before being hidden giving you an opportunity to re-upload. Right now it’s hard to know what exactly to re-upload, so we have actively being working on minimizing the number of times things fail - this should have gotten much better in recent months.
If a capture is in “Processing” it won’t be shown in the feed yet, and it could complete successfully (most cases) or fail (a small number of cases) - not sure if I follow your question though.
So what failed? I cannot tell any difference between this sequence and any other except the “Failed” label. If I just reupload it, what happens to these images and detection - are they actually deleted at the 2 week mark (I guess I’ll see that in a few days)? Should I delete it first? Will I not be uploading duplicates?
red in the feed. Now, it just says 
if you want to re‑upload a failed sequence but you have to wait some ambiguous amount of time to consider a sequence failed before you can re‑upload for sure. It is just a waste of time and very inefficient for any contributor’s workflow. It would be nice to have some sane feedback here.
What? So, it isn’t gone? 
Why do I have 
