DJI Air 2S creates unknown camera calibration images from 5.4k video

[NNaglich]

I have recently purchased the DJI Air 2S which takes 5.4k video. When I process the video to images it shows them as autocalibrated (Unknown 5472 x 3078 px). When aligning the images it's always wrong, even with a "Advanced" everything on high settings (which takes forever..)

Is there a modification file for this camera that someone has already made? I don't understand how to modify it myself.

[cam3d]

Hi NNaglich - I'm not sure about that particular camera calibration, but if you are able to share the images with us through support@3dflow.net we can have a look into what is going wrong and give you some targeted feedback on how to resolve the issues you are having. Aside from this, it's not recommended to increase every SFM setting to high as you may be introducing false positives which result in a dissatisfactory result, and you'll definitely increase processing times, and in most cases this is not necessary.

The easiest way for us to receive image data is through a public download link through a service such as Google Drive or WeTransfer.

[NNaglich]

Hi NNaglich - I'm not sure about that particular camera calibration, but if you are able to share the images with us through support@3dflow.net we can have a look into what is going wrong and give you some targeted feedback on how to resolve the issues you are having. Aside from this, it's not recommended to increase every SFM setting to high as you may be introducing false positives which result in a dissatisfactory result, and you'll definitely increase processing times, and in most cases this is not necessary.

The easiest way for us to receive image data is through a public download link through a service such as Google Drive or WeTransfer.

Ok, thanks! I will send an E-Mail.

[NNaglich]

Hi NNaglich - I'm not sure about that particular camera calibration, but if you are able to share the images with us through support@3dflow.net we can have a look into what is going wrong and give you some targeted feedback on how to resolve the issues you are having. Aside from this, it's not recommended to increase every SFM setting to high as you may be introducing false positives which result in a dissatisfactory result, and you'll definitely increase processing times, and in most cases this is not necessary.

The easiest way for us to receive image data is through a public download link through a service such as Google Drive or WeTransfer.

You were correct in not needing to increase the SFM setting to high. When aligning the photos using the "general" and "default" settings it aligned the photos correctly. So I was able to generate a model with the data set from the DJI Air 2S 5.4k video. Thanks!!

[cam3d]

Sweet! Glad we resolved the issue :-)

[Andrea Alessi]

You were correct in not needing to increase the SFM setting to high. When aligning the photos using the "general" and "default" settings it aligned the photos correctly. So I was able to generate a model with the data set from the DJI Air 2S 5.4k video. Thanks!!

Additional suggestion: since you're using a video, there's no exif data in your images that zephyr can leverage to know beforehand which camera it is woking on, so it needs to autocalibrate the camera each new reconstruction (as in, every new video you convert to frames).

Once you manage to make a good reconstruction (ideally with a significant amount of images from very different angles) that you can confirm has been done correctly (so make sure there's no orientation issue, use control points around the scene if possible to verify that) I suggest exporting that camera calibration (you can do that manually or save it in your camera calibration manager) so that you can re-use it in your next project.

Keep in mind changing focus or settings however (even just the resolution) will require a new camera calibration. However if you're not touching the camera settings, it's a good idea to re/use the calibration so that the zephyr can be slightly faster in the SfM phase, but most importantly, you're helping zephyr in the future for datasets with less images or that are more difficult by nature (fewer photos, more difficult subject).

[NNaglich]

Additional suggestion: since you're using a video, there's no exif data in your images that zephyr can leverage to know beforehand which camera it is woking on, so it needs to autocalibrate the camera each new reconstruction (as in, every new video you convert to frames).

Once you manage to make a good reconstruction (ideally with a significant amount of images from very different angles) that you can confirm has been done correctly (so make sure there's no orientation issue, use control points around the scene if possible to verify that) I suggest exporting that camera calibration (you can do that manually or save it in your camera calibration manager) so that you can re-use it in your next project.

Keep in mind changing focus or settings however (even just the resolution) will require a new camera calibration. However if you're not touching the camera settings, it's a good idea to re/use the calibration so that the zephyr can be slightly faster in the SfM phase, but most importantly, you're helping zephyr in the future for datasets with less images or that are more difficult by nature (fewer photos, more difficult subject).

Awesome Thanks! That helped a lot with a new building I just tried scanning. It was giving some bad alignment with a few photos, but with the settings from another scan it works great.