We should start to "help" ODM

Feedback Ideas and Proposals
1

I want to share something with everybody.
This review focuses on application of UAS in landscape ecology and the questions linked to ecology. It does contain interesting information on Web ODM.

Villarreal, M.L., Bishop, T.B.B., Sankey, T.Ts., Smith, W.K., Burgess, M.A., Caughlin, T.T., Gillan, J.K., Havrilla, C.A., Huang, T., LeBeau, R.L., Norton, C.L., Sankey, J.B., Scholl, V.M., von Nonn, J., Yao, E.H., 2025. Applications of unoccupied aerial systems (UAS) in landscape ecology: a review of recent research, challenges and emerging opportunities. Landsc Ecol 40, 43. Applications of unoccupied aerial systems (UAS) in landscape ecology: a review of recent research, challenges and emerging opportunities | Landscape Ecology

I will not cite, since the details are in the publication:

  • not all studies do use photogrammetry, some focus on questions that can be answered with different methods
  • research lacks implementation and use of open-source software
  • Web ODM is only used for 2 studies of the 161 examined ones
  • there is a lack of reproducible processing code (basis of scientific work is to make results reproducible), since people work mainly in GUI driven software

My suggestion is the following: You have to help write “landmark papers” that use ODM. The science community follows the path of least resistance and the path of what has worked in the past or for others. Currently, Web ODM is not part of this. And by working with actual users (open-source often does everything from a developer standpoint) we will gain much better insights on what functionality and what help people do actually need.

I sadly can not help with this since I am bound up with work myselve. However a strong and motivated group can and will be able to vastly improve the situation.

TO START (suggestion)
How about a video series explaining every detail a user will need to know. Explain every possible setting in the config of a processing job and show examples of the use and misuse as well as best practice suggestions. When doing this do not brush over anything and cover everything as if it was a full lecture. I am thinking about an old Hak5 approach, where different hosts guide though things, with their own projects and work as examples.

The UAV space is getting more closed down and will continue to do so. Regulatory changes, data security and technology advancements will be handled a lot better by large companies than small ones. And these companies will make sure to secure their market position. - This cange will not happen on it’s own, Web ODM needs help.

2 Likes
2

I’m game to help make a series to help people use ODM. Of everything I’ve used, it’s by far the best, not because it gives the best results, but because it’s not a blackbox that I’m expected to trust the results of.

2 Likes
3

Fellow scientist here with a love for open source workflows and free software. I donate regularly to the projects I depend on the most, or I buy swag, such as t-shirts or the ODM book. I believe this is always the most significant support for projects.

That said, in my science, I still rely on Pix4d and metashape, mostly because the WebODM workflow is too clunky for me and would be impossible to explain to my students, who often have limited computer skills and patience.

The biggest drawback right now for me is the GCP interface. Even the paid one is difficult to use compared to Pix4d, which, in my opinion, has the best interface.

Second, the settings are unorganised and all over the place. This could be structured in a clear workflow by simply adding a few headings in the settings tab. Again, Pix4d has a very simple to understand 3 step structure with simple settings and more advanced ones. Opening the settings in WebODM is just overwhelming if one opens the app only from time to time.

Third, support for larger datasets is tricky. Again, the infrastructure is there, but for the layman not doable.

Fourth, the output would be much easier to deal with if it were just to appear in a project folder structure. Regular users understand that, but may struggle with the web app concept where you download your outputs to your computer, when they should already be there.

My biggest recommendation to the project is to write a proper desktop app and rethink how the user can manage the settings for various steps in the workflow. The developers seem to come from a web app and server background, hence the browser interface and all the docker stuff. But especially the browser interface seriously limits the interaction with the data and workflow. At the moment, I find the GCP interface incomprehensible and impossible to teach to a student who could potentially use the app.
Alternately, as has been suggested before, make it a plugin for QGIS, but I am not sure how much that would simply the task.

Of course, i am really grateful that the project exist and would love to see it reaching the next level. The backend is there, but the interaction with the user is what is the limit in my opinion.

3 Likes