Try New Video Format Support (& ExifTool) in the the Desktop Uploader Beta

Hi all! We would like to invite you to participate in the Mapillary Desktop Uploader Beta program! This is a great way to get access to new features and capabilities and an opportunity to share your feedback with the Mapillary team.

We’re starting the beta test for a big new update which brings ExifTool integration to the Desktop Uploader and enables support for uploading images and videos from a huge variety of cameras such as BlackVue, Anker, Vantrue, Viofo and many others. If you were unable to upload to Mapillary due to our limited camera support before, or had to do complicated workflows to convert videos to images, now is the time to try uploading directly! There are also several performance and stability improvements as part of this release.

To submit feedback, post in this forum thread with example videos or photos that you’re having trouble with.

You can download the latest beta version for Windows, Mac & Linux.

Once installed, the beta version will be called “Mapillary Uploader BETA” so you can keep using the public version along with it. The beta will update to the latest version automatically so you’ll always have access to the latest features in development.

Thanks for your support and happy uploading!



Bumping this up - has anyone tried this new desktop uploader?

We’d love your feedback on videos/cameras that you have tried and what your results were.


A&B Courier will be resuming Front camera capture with Blackvue 900-S shortly. Previously we where uploading via Cyberduck software to an S3 Mapillary bucket. We will be sure to provide feedback on the new application.

Thanks for being Mapillary!


Love to hear it - thank you Eric and looking forward to any feedback :slight_smile:

Full Exif support in the Desktop Uploader, in other words the capability to modify Exif tags (automatically) in a UI before upload, sounds really nice indeed. :+1: We have been waiting for this for a very long time.

I have just one nit; I would like to see an aarch64/arm64 release too. :wink:

1 Like

Opening a gopro max video fails with Ubuntu 20.04:

[19:04:20.975] [debug] [mapillary:tools] [111043] Error loading Python lib '/tmp/_MEI1F08gI/': dlopen: /lib/x86_64-linux-gnu/ version `GLIBC_2.35' not found (required by /tmp/_MEI1F08gI/

Same works with the old version.

Since I can no longer download my photos from my device due to Android restrictions. I no longer use the desktop uploader.

So, what does ExifTool support actually bring? So far I only see the exif summary on images, no ability to edit etc.
Or is this a release for people with video workflows?

1 Like

Im waiting for the desktop uploader to be able to remove duplicates when stopped or to slow. Until then no sense for me. Thats kind of basic function. Dont know why its not there yet.

Second thing. I was trying to record timelapse video 360 with gopro max but GPX position was so off that i was forced to delete everything.

@masterofnoroad I had the same exact issue two-three weeks ago with my GoPro Max. I probably have to recorded two days of capturing again, or manually edit all gpx tracks. The Google Street View community on Facebook were talking about some unusually strong solar winds this month causing GPS-devices to suffer badly. It would be interesting to here if anyone else had problems with different cameras.

Im waiting for the desktop uploader to be able to remove duplicates when stopped or to slow. Until then no sense for me. Thats kind of basic function. Dont know why its not there yet.

Because it is not at all trivial to implement. Indeed, numerical equality is easy to implement and to define. However, it is very difficult to define and even harder to implement contextual equality, especially with regards to images. When are two images identical or equal for you? When they have identical GPS positions in the metadata, or maybe when GPS positions are just close enough? What is “enough” for you or somebody else? Or, do you consider images the same when their content is visually identical, and if so is weather part of that equality?
Furthermore, two JPEG images can be numerically identical in the visual space (decode into identical bitmaps) and yet be differently coded resulting in two different files. So, you see, it is not as trivial to detect duplicates as one might initially think.

1 Like

The uploader now skips images which have an exact same latitude, longitude and compass angle as another image in the session. This should prevent duplicates when standing still (e.g. stopped at a stop light).

We have customizable distance-based duplicate checking in mapillary_tools: GitHub - mapillary/mapillary_tools: Command line tools for processing and uploading Mapillary imagery Would adding this setting to the “Editing” section help?

1 Like

ExifTool support allows us to read EXIF data from many cameras which we couldn’t process previously (mostly videos). This enables upload automatically, without additional need for manual tweaking.

1 Like

Thanks for confirming, I think me (and some others) have mistaken this for some sort of advanced EXIF editing being added

1 Like

So far I only see the exif summary on images, no ability to edit etc.

:thinking: @boris I too was actually hoping for the capability to edit Exif tags per sequence and per image (in a convenient UI). For example, my camera produces quite some amount of useless data in vendor extension tags, which nobody else than the vendor can meaningfully interpret. Ever since I started contributing to Mapillary, I skim my image files from this additional ballast. At the same time, my camera does not automatically add the GPSImgDirection and FocalLengthIn35mmFilm Exif tags. Sometimes, I forget to reset daylight saving in the camera, so that local timezone timestamps are off. So, I need to fix these broken timestamps before upload. Thus, it would be nice to be able to edit or fix some Exif tags automatically in the Desktop Uploader. Sure, I do understand that different cameras may require different additional functionality. However, being able to modify Exif tags just manually (per sequence or per image) would be a step forward.

Thanks for the feedback and feature request @GITNE! That makes sense. The approach that we’re following is trying to fix these issues automatically (because as you mentioned different vendors have different tags and not all users will go through the hassle to fix the issues even if there was a nice UI for it). Here’s what we’re thinking:

  • for local timestamps we’ve implemented a feature for videos that are uploaded which will automatically correct timestamps based on GPS time (which comes from the satellites and is always correct). Adding the same support for images is on our backlog.
  • for image direction we are also planning to improve our structure from motion detection to correctly set the direction even if its not set on the source video or image files.

Until these things to live I think running exiftool manually remains the best course of action. Thanks again for all the feedback and contributions! We really appreciate it.


Thanks for reporting! @tao FYI - maybe something about the older version of Ubuntu? @eserte I think Ubuntu 20.04 was released a couple of years ago, are you able to try updating your Ubuntu install to see if the issue is resolved?

I think Ubuntu 20.04 was released a couple of years ago

However, it’s an LTS version and will be supported until 2025. It would be good if the uploader would work on supported OS versions.

In real life there are no photos with EXACT same lat/lon as far as you use standard gps receiver in camera like gopro so preventing duplicates with exact same coordinates serves only protection from double uploading same files.
I would like to have in mapillary desktop uploader customizable distance-based duplicate checking as in command line mapillary tools. My usual parameter is to delete photos spaced less than 2m and it covers crawling in high traffic or stopping on lights or short stops when doing bicycle trips.

1 Like

I think its more a problem of video timelapse mode. When Im using photo timelapse its working quite OK. Unfortunately its only possible in 2s interval but for bicycle its enough

1 Like