Insta360 double Fisheye DNG's

[wizprod]

Hi,

Is there a way to import dual-fisheye DNG's from the Insta360 RS One 1 inch?

I can of course get it to stitch after converting to 2:1 ratio equirectangular images, but hoping I can better image quality without heavy image processing step in Insta360 Studio.

[JMR]

Not sure with insta dng's but I've had some success with ricoh Theta and photoshop.
Import DNGfile, crop to the left fisheye half extent and save as jpg, then import again and crop to the right half to save as new jpg
just make sure the left and right cameras are always exported into different folders and you let Metashape know they are different cameras. Use the photos as multicamera or as single camera files, its up to you.
Regards
José Martínez
Geobit & Accupixel CTO
Metashape training providers

[wizprod]

Thanks!
Do you have an automated way to do this? This would be a pain to do for 500 shots

[ilia]

Hi wizprod,

Do you have some sample data? I would like to try it. If it works I will share with you how to set it up in Metashape.
What Metashape version do you have? Standart or Pro?

[Alexey Pasumansky]

Maybe it would be helpful to take a look at the following video:
https://www.youtube.com/watch?v=fwjS1CdGl5Q

Author suggests (at about 24 minute) to use the frames from each camera separately and split the calibration groups, also to set Fisheye camera type prior to the alignment process.
Video also includes the control and check points for the accuracy validation.
If interested, you can scroll to the end when proper processing approach is suggested (https://youtu.be/fwjS1CdGl5Q?t=1440) to check the results and error values on check points.

[wizprod]

Thanks for the comments, much appreciated.

I will see if I can generate some non-sensitive, photogrammetry suited sample data. In the meantime, You can find some DNG's here: https://drive.google.com/drive/folders/1TACxCBcZrJ2P-qQNLdbzphU_5G0CVjcB

I managed to make ChatGPT 4 make me a python script, that would split the images into seperate folders, so Metashape would easily import these as a multi camera project. After trying to figure out pixelpitch and focal length (this was a pain, with little to no hard spec on the sensor), I managed to get a very decent result, where some manual markers had 0.3 px precision as compared to the 3 px precision when I was using the equirectangular images made with Insta360 Studio.

So far, so good!

[ilia]

Dear wizprod,

I've checked your data and it seems that to have your data recognizable and actually using both cameras simultaneously you will need at first slice each image in two.
Otherwise on a single image you have two optical centers, which are not manageable but most of photogrammetry pipelines in the world, including Metashape.

And also to make sure it actually works you will need some dataset with different camera positions. Like 6 and more.

[wizprod]

Slight update on this: I imported the 2 cameras (after splitting) as a multi-camera system, which have proven to work nicely, as it is indeed a very rigid setup. Using some GCP's, it is stitches very nicely.

[radar_boy]

I'm just starting out with the same camera. I see you have a functioning workflow splitting the images, but I was wondering if you'd compared the output from this with output using the rectilinear images processed by Insta360 Studio please?

[luisSeyfer]

You can use xnView to split. https://www.xnview.com/en/xnviewmp/
Tools > Batch Convert, add action: Crop

[antoineB]

Maybe it would be helpful to take a look at the following video:
https://www.youtube.com/watch?v=fwjS1CdGl5Q

Author suggests (at about 24 minute) to use the frames from each camera separately and split the calibration groups, also to set Fisheye camera type prior to the alignment process.
Video also includes the control and check points for the accuracy validation.
If interested, you can scroll to the end when proper processing approach is suggested (https://youtu.be/fwjS1CdGl5Q?t=1440) to check the results and error values on check points.

hello Alexey,
METASHAPE v2.2.2 ( build 21287 ) is reporting that issue when importing directly INSV file format in video importer

Error: Can't read video: La demande est incorrecte dans l'état actuel (0xC00D36B2): C:/Users/VID_20251007_105403_00_015.insv

do you have any idea ?

As a rule of thumb, generally, with insta360 product, just changing *.INSV extension to *.MP4 make it avalaible for parsing the video.

[Alexey Pasumansky]

Hello Antoine,

Can you please share such INSV file to support@agisoft.com that fails to import with default extension but works after renaming to MP4?

I have checked the Import Video command on some random INSV files downloaded from the Internet and it worked both with and without renaming.

[Alexey Pasumansky]

Antoine, please also specify, if you observe the same behavior for all INSV files that are captured by your cameras, or the problem is observed only for the large files above certain file size?

[antoineB]

Antoine, please also specify, if you observe the same behavior for all INSV files that are captured by your cameras, or the problem is observed only for the large files above certain file size?

same behaviour and console error for all those files showed in that attached screencap

[Alexey Pasumansky]

Hello Antoine,

I have checked the file that you have provided. Seems that it is using HEVC codec that is not available by default on Windows 10 and Windows 11. So cannot even play the video by Windows 10/11 default player (renaming to MP4 doesn't help here). Metashape fails to import the video by the same reason.