Using Parrot Bebop 2 fisheye lens to model

Has the Parrot Bebop 2 camera been accepted into WebODM? I tried using the fisheye lens option and have had zero luck. Thanks

It should work; perhaps the images don’t have enough overlap. Have you followed the best practices as indicated on ?

1 Like

I’ve made sure to run my flight with 85% overlap, just to test this theory. It does not work at all. If I rerun the processing with the exact same settings, I get completely different results every time. Maybe ODM truly does not support the bebop 2?

Following up on this. I’m still having no luck with WebODM. I would prefer to use this over Pix4d, as I can run it on a local server.

We got support for the Bebop2 camera merged in upstream OpenSFM a while back, and so it should work in WebODM that is semi-recent.

Are you fully up to date?

Can you share sample data that isn’t working for you?

1 Like

Are you still interested in a Parrot Bebop 2 dataset? I flew a circle around a building in a parking lot and WebODM hasn’t been able to construct a useful mosaic from those images. However, I had much better success using WebODM after I stripped all exif data from the images. There is still some distortion, but there is also a building in the mosaic and in the 3D data! I haven’t been using WebODM very long so I’m probably missing some detail about how to configure the processing but somehow the image orientation information isn’t being read or used correctly. Overall I’m very impressed by WebODM and the underlying programs and I’m slowly tiring to understand how it works. I do a lot of underwater mapping with fancy sonar and working with the drone images is much more complicated!

Where can I find guidance about how to upload my parking lot data set, with and without the exif data? – Thanks!



Sure, test data is always great!

We like our website for storing and sharing sUAS data.

I;ve uploaded the two data sets (with and without exif data) to DroneDB. They can be found at:

Hopefully you can show me how to make sense out of these bebop 2 images!


Thanks for these!

Looks like we may have a processing regression!

Auto doesn’t detect it properly, and even forcing fisheye is not sufficient (which should work with the patch I put upstream a while ago):

Might need YanNoun’s assistance on this one.