Trying to understand settings and GSD of outputs

Hello,

I’m new to ODM and this is my first post. So first off Hello, and looking forward to being part of the community.

This post is exclusively referring to WebODM.

My issue is that I am not getting expected GSD results for my orthophoto based on my WebODM setting. The attached table is a synopsis of each run (all on the same imagery set) and maybe someone can help me understand where I’m misinterpreting my settings.

I’m using a Pix4d run as the base line where the orthophoto GSD was 3.17cm. The “Ignore-GSD” is slightly ambiguous to me. The way I’m understanding it is that if it is checked, it ignores the internal calculation of GSD and uses a user supplied GSD or the default of 5cm if not specified. If it is not checked it will internally calculate the average GSD which will then be reflected in the output orthophoto. With this understanding, I anticipated that not checking the “Ignore-GSD” setting would yield as GSD somewhere near ~3.17cm. However, when I use the “Default” template and run it without touching any settings and then do a new run only changing the “ignore-gsd” to checked it yields identical results.

Another bi-product issue I noticed was that if I download “All Assets” and unzip it the resolution of my orthophoto is different (closer to expected value) than if I download orthophoto alone (See spreadsheet for numbers).

Could someone please help steer me in the right direction? The biggest thing I’m looking for is ODM’s ability to get the GSD close to the anticipated value without me having to rely on prior knowledge (Pix4d results) and supplying it!

Thanks!

ODM_Results

Welcome to the community! :hand:

Using --ignore-gsd will disable an optimization which optionally resizes images and/or caps the maximum output resolution of 2D results (orthophoto included). In short, this is to automate the guesswork of computing the “highest” resolution possible, you can set --orthophoto-resolution to 0.001 and the GSD estimate will choose a more appropriate value for it, depending on the dataset. So if you want control over the resolution of your file, you should definitely set --ignore-gsd (which can sometimes overestimate the true GSD value).

The difference between the orthophoto in all.zip and the one you download directly is that the one you download directly has been reprojected to pseudo mercator (https://epsg.io/3857), whereas the one in all.zip is in UTM. I’m guessing that’s why you see the discrepancy.

I’ve debated whether we should always return the UTM copies to avoid confusion and perhaps the program should be modified to do so. (Opened https://github.com/OpenDroneMap/WebODM/issues/800)

Piero, thanks a lot for the quick response. I was missing that crucial combo of having “ignore-gsd” unchecked and setting a super high resolution in order the let the process dictate the estimated/average GSD and apply it, but it totally makes sense now. I did not understand that the default 5 cm took precedence over the calculated GSD if it was a higher resolution. My updated table is as follows:

ODM

As far as the difference in projection, I personally don’t have a strong opinion either way but often prefer the UTM projection, but regardless there is resolution degradation happening in the reprojection. For that reason alone, I would definitely prefer UTM if it meant avoiding the loss in resolution.

Thanks again for all the help and quick response!

2 Likes

Resolution Difference seen from the same run:

ODM_Resolution

The dimension (width x height) of the two files should be the same, so you shouldn’t have a resolution loss. Make sure you are not looking at an overview (that’s also a difference between the two files, the UTM one has no overviews, the pseudo mercator one does).

I attached a screenshot of the data I’m pulling just to make sure we are on the same page.

ODM_Files_Downloaded

When I pull both of these files into QGIS and look at the properties, this is the resolution that I have recorded in the above tables which seems accurate to what is being displayed.

Image properties are not the same for the two shown above:

ODM_ImageProperties

Mm, I stand corrected; this could be a bug then. :pray:

Would you be able to share the UTM version of the orthophoto (even privately)?

1 Like

PM sent. Hopefully it is not to big of a bug to fix. Based on the table I provided, the direct othro downloads have some uniformity to the GSD’s. It looks like when it reaches a certain threshold the GSD doubles… Don’t know if that helps at all.