I have a gps track with 5 points per second. Would that work with Mapillary?
I guess so I walk and use a insta360 X4 and get my GPS track from the app Runkeeper…
My workflow
Lesson learned
- looks when I walked around a cemetery with circles I need to walk slower i.e. I should maybe then have mor points per second… when walking a trail its no problem
Is off-topic but I wonder why horizon levelling looks so bad. Did you not set Stabilization type to Motion when exporting?
1 Like
-overwrite_original -m
Very helpful. Thanks
Here you can read the recommended settings for driving, cycling and walking: https://help.mapillary.com/hc/en-us/articles/11951588568604-Insta360-X3
@Hol_Ger why horizon levelling looks so bad
yes that is my problem… still learning - issue…
when I compare to @brunofilmklov who is doing Mapillary in my near he has another quality level and another camera…
It is not the camera. See Mapillary an even Mapillary.
You can pan around even the last image and the horizon levelling looks really good.
Try to switch on the camera only when the stick is in perpendicular position.
1 Like
Thanks @Hol_Ger I will do some walking today
Hi Salgo60,
Image Orientation
Yes, though my ThetaX camera doesn’t have a Horizon Lock feature, so I have to do that the old fashioned way* - by attempting to ensure that it’s leveled at all times 
.
Though it does have auto orientation, so (as @Hol_Ger mentioned above) the orientation when the the recording button is pressed does matter in my case; (there have been several occasions where I accidentally started the a time delayed recording at close to 45° angle and corrected the lean and then noticed failures processing failures, as the camera had locked in the rotation and tagged the recording with a 90° rotation…)
I believe this can usually be corrected in post processing by using the “–orientation=0” option when processing the sequences for upload, but once uploaded there are no tools in the app to manually correct the issue.
Position accuracy
Most GPS receivers can only only produce one position fix per second, so anything in between these point usually has to be interpolated during processing. So yes you really do need to mind how quickly you change direction; the slower you move, usually the better the accuracy of the location.
Further, the GPS in my camera doesn’t like being moved around randomly, so when holding it on top of the selfie stick, the accuracy was not good. Most smartphones have the same issue, e.g. if you hold it in you hand, the small random movements the precision can quickly degrade from less then 3m to more than 15m. In my experience placing the receiver it in a pants pocket while walking provides better accuracy, then holding it in the hand while standing in the same spot, though still not nearly as good as when pressed against the upper-body/shoulder. (The double selfie-stick above seems achieves similar reduction in random movement as well.) Vibrations through, seems to have less of an affect; while bumps can make the GPS severely overestimate the actual movement. (Tall buildings nearby can also affect the accuracy quite a bit.)
Thus if you are unhappy with the accuracy, I would recommend also reviewing how you are managing the GPS tethered to the camera, as that could help improve the location pression significantly.
GPS track usability
Regarding the original question about the GPS track usability - 5 points per second is more than most of us get, so if saved as a .gpx file it should be usable. That stated, Mapillary recommend against using a separate GPS track, as it can be quite error prone to align the timestamps in the GPS-track with the time stamps with the images/video.
(Yes, it seems simple, but also in my experience it is very easy to get the offset wrong without noticing in time… so I personally view it a as a last resort, to be used only when the it in not practical to redo the image-sequence/video-recording with the correct GPS position data embedded.)
/Bruno
PS.
Orientation stabilization
For walking I tried using just the Theta TM-2 stick, but the results proved unsatisfactory. Adding a second insta360 sick below, and a 200g weight at the bottom improved the stability significantly, and ensures the stick will be hanging plumb unless it is pushed into an angle. Drawback is that the long stick becomes fragile and can easily be bent…
Secondly the TM-2 stick has a fixed relation between the bottom of the stick and the camera - as there are only two ways the camera can be mounted, and this helps in visually confirming that the camera is consistently pointed in the direction of movement, while held well above the head.
DS.
2 Likes
With X4 this problem is avoided by starting the recording not with the button at the camera but with the Insta360 app. Then the camera’s time is set accurately. See Operation Guidance - X4 Support. Thereafter I disconnect the smartphone from the camera to save smartphone’s battery power.
@brunofilmklov What is your Mapillary user name to see some of your tracks?
1 Like
He is bruno360 does both car, biking and walking Mapillary I use them to look for BBQ places in Sweden and add them to OSM and grillplatser.nu 
1 Like
Nobody is interested in recordings where the vehicle is currently in a traffic jam; Mapillary obviously recognizes such sequences based on the GPS position. This is a good solution for videos that contain GPX information. Anyone who is currently recording video and GPS tracks separately must make sure to interrupt the image recording immediately if the vehicle comes to a standstill. Otherwise, the GPS track continues while the vehicle is still in a traffic jam. It would be ideal if you could optionally switch off the function that Mapillary skips GPS sequences when the vehicle is stationary. The alternative is to find a solution to merge GPS and image using independent software before uploading to Mapillary.
I record only hikes. So no problem with traffic jams
1 Like
Even when hiking, it happens that you have poor GPS reception between the trees. Separate GPS devices often have better performance. When later combining video and GPS in the desktop uploader, it is difficult to synchronize video and GPX, especially on hiking trails without a fixed landmark such as a curve. I would also like to point out that if you stay at one point for a long time, Mapillary will skip these GPS points while the independent video continues. This means that the GPS movement does not pause and continues hiking while the video still shows the smoke of the campfire at the rest area.
1 Like
I had these problems before using the insta360 app to start recording. Maybe it is less error prone with intervall photos as I do them than with video.
Never realized such a problem when I met someone and talked a while but forgotten to stop taking photos.
Typically yes; and that is correct. (And in the context of hikes, you can just substitute “traffic jam” with “campfire” - its non-movement in both cases.)
Hm, i think you intended to state “GPS information” (alt. GNSS information), as .gpx (text) file-format are used when the sequence of position+timestamp are in separate file. (You can use for example exiftool to create a .gpx files, by extracting the GPS information from a video file, for example to uploaded the path you have hiked to osm.org…)
Technically this is is true, however GPS track will just indicate that you were not moving - so not an issue. The real problem is that the video data will needlessly consume more storage space, and thus you will storage soone then if you stop the recording. So good practice, if you believe traffic jam will last longer than a few minutes. (Drawback is that you might forget to start the recording after the traffic jam has been resolved - been there & done it 
and expect to do that mistake more again.)
(Though if you are recording the smoke of the campfire at the rest area (and your companions), storage space usage might not be a concern – as the video probably have other uses then a upload to Maillary
Not true, the mapillary actually using those GPS points to skip some of the images in the sequence - thus cutting out the “uninteresting” bits automatically.
If you disable both the duplicate image and the sampling distance functions, then you would be uploading a large number of images from within a ~5cm/~2" radius everytime you are stationary in “traffic jam” or by the campfire. As you noted initially, is very little use/interest for nearly identical images on Mapillary, so best not to disable these functions. [The duplicate image function requires minimum distance of 1dm for image sequences. The sampling distance function is only applicable to videos, and defaults to 3m.]
Possible, but it’s basically with the “video_process” command does, extract the images and add the GPS info; though you might need to use for example the --overwrite_EXIF_gps_tag option get what you want… (BTW, for 360 videos some of the metadata will be missing in the extracted images, so you might need to use for example exiffixer before uploading the extracted images to Facebook.)
Basically what the Insta360 app does, is to update the time on the camera just before starting the recording. Thus if you have a GPS logger on the smartphone, then the timestamps in the video and the .gpx file should match up perfectly.
The typically problem is that the camera clock isn’t set correctly in the first place – or not perfectly calibrated and have slowly drifted away from the true time; and thus you need to compensate for the difference in timestamps created on the camera and on the GPS-logger - and as you noted this can be difficult and error prone.
That highly depends on GPS devices used, and what kind of terrain you are hiking in.
If you are hiking close to tall buildings (e.g. central Rome), or near cliffs, then you need something better then the standard GSPs receivers (found in most cameras/phones).
I have tried connecting a simpleRTK2B GPS receiver to the phone (used as the GPS-logger) a few times, and while the precision was improved from around 3m to ~0.3m. Unfortunately it tended to suffer poor GPS reception between the trees and stopped working entirely if there were too moisture in the air increased too much - (probably didn’t help that my attempt was done in the late autumn - just before the a drizzle or snowfall started).
Don’t think the receiver was bad, it was just a janky setup; in the end it seemed easier to just ensure that the GPS receiver in the camera worked as well as possible, and th get good enough position data for my needs.
An example of the situation I’m talking about: I drive off and stop at the barrier at the exit of the parking lot and unfortunately have to add more coins. However, the GPX track continues unchanged through the barrier, and then the GPX track is uncatchable ahead, even though I started the video and the GPX recording at the same time. /Video and GPX uploaded separately using the desktop uploader)
GPX and video uploaded separately in the desktop uploader example
Addition: Mapillary managed the situation at the barrier, but not the crossing pedestrians, the GPX Track then runs ahead of me
Yes, see the problem.
- Yes, I also believe you got the time synchronization wrong - you need to re-adjust the time offsets, so the /Video and GPX are synchronized. It probably only off by something like 30 seconds, but that translates to significant distance.
- There seems to be some false movement, especially around barrier. Maybe you accidentally took the GPS receiver with you when you add more coins ? Or accidentally bump the receiver? (I have seen similar false movement happen to me, when I have stumbled or bumped into the selfie stick - and then it can take a while before the GPS seems to complete recalibrates and drops what seem to be over-compensation from the accelerometer input …)
However this could as easily be a cased by the faulty time-synchronization issue.