yeah, i noticed some of my longer sequences being cut up and only a small part appearing online; like this one f2s9G4uIO7Bkd89UxpwUiu among many others, i tried uploading the remaining images again (for KeGjiGPtpYerBneiHR00bu as example) but that also didnt work
I guess this approach is basically flawed, whether itās just one image ore an entire sequence. I often take images that donāt look like ānot-street-levelā images, but they are useful for (OSM) mapping purposes. Examples are names of companies on signs or even door bell labels, or text that complements a traffic sign, or even city-limit signs (which in Italy are rectangular black on white and look, without context, like any kind of text on paper), or renderings of developments that are useful to map construction sites.
Can we not have the bot hiding a suspect image image temporarily from public view, but flagging it to the user with a working URL and let the user decide whether to render it visible or not.
It is certainly not acceptable having a bot in the background hiding images without warning the user about it and without giving the user a chance to interact.
(I am waiting for many hundreds, or more, of images from 8 Nov 2020 onwards)
4 Likes
Unfortunatelly whole sequences are still being flagged wrong. My whole 27th march data went there, so 100% pictures were flagged wrong. And that is something new. Usually its just one sequence. Maybe its due to me upgrading to new mapillary tools? Or because I started to crop all pictures to get rid of a car hood in front that is taking 1/5 of the picture?
Hope You are still working on the solution and the picures will eventually show up.
My latest set of about 2000 images have not appeared on the map nor in the right-hand pane under āFeedā and āUploadsā. I donāt have anything showing as āpendingā in my āFeedā. Because I am worried about images disappearing, I only upload a set of images when the previous one appears on the map, which means I have a large backlog.
The latest images are from 08/11/2020 (i.e. 8th November). Example sequence IDs: 71dbd52b-fea1-40d4-8612-0b113a6bfd84, 7d350271-e44b-4664-96ef-0252f1e5aa0e
I was exaggerating a bit. Sure that a site with the visibility of Mapillary needs filters to protect the site owners and also the contributors.
On the other hand the ease of use for the contributors should not be underestimated. For years I uploaded my images directly from the smartphone, without keeping copies locally. The site also had the nice feature hat I could download my own unfiltered and unblurred images.
Now what about a system that puts the rejected images in a bin from which contributors can download only images that they own. And complement that with an automatic message to the contributor that some of her images have been deleted. You can also put a time limit in that message after which you may empty the bin.
Obviously you need a system that is automatic on the Mapillary side and any work needs to be done by the contributors. And then you have to clearly state what we are not supposed to upload.
3 Likes
Good news
We are doing a complete backend overhaul that will change the way images are processed. By the end of the month you will no longer need to worry about the bot flagging your images and holding them in limbo.
Images will be shown by default like they were in the past. In future, we can work with you all to determine what flags might need to be in place if there are inappropriate things being uploaded. Iām happy to say that 99.999% of images are relevant street-level imagery.
4 Likes
Thatās great news!
I will happily work in the future to help create some way to filter out unwanted bad imagery. Still it need to be half-manual process or at last with some fail-safe.
Hope it will work backward with already uploaded data. Waiting for the update to take place.
1 Like
Same here, my sequences are divided by 300 for no reason.
If you are capturing with iOS app, there is 300 images limit per sequence. Longer sequences will be divided into shorter sequences after processing.