I’m trying to upload a huge amount of images captured with Insta360 Pro, but Mapillary says that the data doesn’t have geotags. When I look into the EXIF of the images, all the photos are geotagged. How can I upload the data? Mapillary suggests to geolocate manually the images, but given the amount of photos it’s not a realistic option to me.
We have found the issue regarding the Insta360 Pro problems when trying to upload data to Mapillary.
Insta360 STITCHER software exports data as JFIF, and it seems to be the reason why Mapillary uploader doesn’t recognize the Exif information. We noticed that when we used ACDSee software for export the same images. Once this step was made, the images where downsized and correctly displayed in Mapillary uploader.
Given that there’s any change in the Exif information, we inspected the same image exported by Insta360 STITCHER and later by ACDSee using the Notepad (dragging and dropping the image in a Notepad window). Here’s how the information looks:
Insta360 STITCHER: ÿØÿà JFIF ÿáNExif II* >
H i‡ V %ˆ À Insta360 Insta360 Pro †’a W h charset = Ascii {“info”:{“gyro_stabilized”:false,“initial_view_changed”:false}} N W & > e àe † g —f † Ý<
ÿáhttp://ns.adobe.com/xap/1.0/ <?xpacket begin='' id='W5M0MpCehiHzreSzNTczkc9d'?>
<x:xmpmeta xmlns:x=‘adobe:ns:meta/’ x:xmptk=‘Image::ExifTool 10.10’>
ACDSee: ÿØÿá Exif II* ’
› ¨ e ° ( 1 ¸ 2 Í i‡ á %ˆ n Insta360 Insta360 Pro ð ð ACDSee Ultimate 2018 2018:07:27 10:50:12 †’a W ’ 660 charset = Ascii {“info”:{“gyro_stabilized”:false,“initial_view_changed”:false}} N ¼ W Ô ì e àe † g —f † Ý<
ÿávhttp://ns.adobe.com/xap/1.0/ <?xpacket begin="" id="W5M0MpCehiHzreSzNTczkc9d"?>
<x:xmpmeta xmlns:x=“adobe:ns:meta/” x:xmptk=“XMP Core 5.5.0”>
We hope this information helps Mapillry to solve the issue related to Insta360 Pro imagery upload.
I know another contributor who is testing the Insta360 Pro. He doesn’t had problem to upload the images, but I don’t know if they have used the python scripts or the web uploader.
Finally we figure out how to extract the GPX info from the Insta360 videos.
The process to extract the GPS track was this:
We used the GPSreader.exe tool included in the Insta360 Stitcher software. The installation path is C:\Program Files (x86)\Insta360Stitcher\tools\prostitcher\gpsreader.exe
With the Command prompt, we wrote this: C:\Program Files (x86)\Insta360Stitcher\tools\prostitcher\gpsreader.exe D:\video.mp4 D:\GPStrack.txt D:\video.mp4 is the path to a 1 frame video generated with Insta360Stitcher and D:\GPStrack.txt creates a text file with the track data
Thank you very much @javiersanp. Unfortunately, Mapillary informed me that they still need to do codec testing and other things to be able to upload videos with the Insta360Pro camera . All this process has been to take a step forward, and two backwards I guess we’ll have to wait a few months until there is support for the camera from the friends of Mapillary.
Sure we can, but given that we have hours of videos and that the information was gathered in a project, we can’t spend time doing all that process (that I’ve already described in this post). We are taking about 18TB of images, with all the ours of processing and uploading that it could take. Certainly that was the first option we explored, and aprox. 15.000 images were uploaded (and that only from a single video of one hour long, and a couple of hours of processing using the Web Uploader).
Maybe in the future we will be able to do that, but in the mean time we were looking to a viable and easy way to share all that information. We can’t storage for much time the data, and we don’t need all the data also. The idea of uploading it to Mapillary was to share information that maybe someone else could find useful, but given the technical issues and restrictions maybe we’ll have to wait.
Many thanks for your help and for the Python script @javiersanp It works great! I think Mapillary people will find it useful too, in order to improve Command Line Tools and support Insta360Pro images and videos, because uploading the images now requires additional work to make them able to upload with the Web Uploader. With Mapillary Tools we couldn’t upload the data, so there’s no doubt that technical support and improvements are needed from Mapillary’s developers.
We tryed both ways @StephaneP But there is too much information and we don’t have enough available storage for the generated data. Uploading the videos seems to us an easier way, given that our project ends in a month. So we don’t have too much time to generate, tranform (remove JFIF), storage and upload stitched imagery. That was why we were looking to upload the videos directly to Mapillary. Uploading images to Mapillary isn’t the main objective of the project, so we just wanted to share data that we know could be helpful to the community.
Many thanks fot all the support and interest. I hope our data could be displayed in Mapillary in the future.
This is a side comment, but that isn’t too much capture density?. 15000 images in one hour are 4 images per second. Keep one image each 2 seconds (1800 images) and discard the rest. This will be sufficient.
What I would do in your situation is extract the images of one video, process, upload and delete the video. Repeat for each video and you will not need to much hard disk space (only that to process one video each time). This could be done in batch mode.
Certainly there is too much overlap @javiersanp . We could generate videos at 1 fps, 5 fps and 10 fps. Unfortunely, Insta360Stitcher is not the best software and it’s very limited, in comparison with other softwares. Sometimes it takes more time stitching at 5 fps than 1. We can’t use third party softwares to composite the videos, given that the gpx information requieres gpsreader.exe and seems that it’s specifically developed for Insta360Pro. Adobe softwares, for example, delet the gpx data.
The ideal way to select images could be develop a script that filter the images by it’s location, specially because Insta360 doesn’t stops recording when the vehicle stops. But as I said, uploading info to Mapillary isn’t the aim of our project, and we can’t spend more time working to solve this troubleshout. We have been trying different ways to upload the data for three weeks, but we have arrived to a deadend.
We’ll supply some videos to Mapillary developers, so they can work on giving support to Insta360Pro videos and stitched images.
The mapillary tools (python) can flag duplicates images, but I agree that it’s not the most user friendly tool. I’ve ask for a GUI on top of the python script years ago.