This software is becoming increasingly frustrating. It simply doesn’t want to process single layer TIFF multispectral images.
I want to use by local computer to do the processing as it’s a pretty decent Mac with an i9 processor and 48gb RAM. I don’t want to use the lightning network node as I have 400+ images. I should be able to process on this desktop.
I select ‘node-odm-1’ which is the automatically generated one; I assume this is the desktop? I can select ‘multispectral’, upload and within 5 seconds I get an error message saying ‘cannot be processed’. Why? A clue would be good.
Anybody out there have any ideas please? Is the node I’m using the desktop? How would I know? Why is there no clear option to process locally? ARGH!!
1 Like
Hi lowaltitude,
Sorry you are having trouble. To be able to help we’ll need some details (and yes, the node that isn’t lightning is your local node):
Sensor, number of images, some details about how you did your flight, and your processing log is a good starting place. Extra points for sharing a link to your images in case it’s a more complicated case.
2 Likes
Usually an early exit for Multispectral processing indicates that not every band in the dataset has the same number of images for the collection.
For example, for a survey of 100 collects on a four-band sensor, we’d need to have 400 TIFFs. If you have 399, 398, etc, that means one or more bands are short and the analysis will stop.
2 Likes
I know this is no help technically but the first thing I tried when the update came out just a few months ago was try to process a few multispectral data sets. Mine on a windows machine (MSI Leopard GL75 I7 10750 and only 32gb ram 1660Ti 6 gb) processed all three faster than previous versions of WebOdm ( I mean over the past three years of versions) and without the indexing errors that we were getting last summer. My data sets are all over 3000 images from a micasense rededge MX.
2 Likes
Thanks for the reply. That makes sense, as the number of images are odd, rather than even. I’ll take another look at it.
Interestingly I was able to process the dataset in a Pix4DFields trial without an issue, so maybe the validation they and WebODM use are different.
1 Like
That’s interesting. I’m new to WebODM and have the latest version. Are you seeing the issue with needing the exact ‘4 images’ to complete processing? For example, 400 images rather than 399 images. I would have thought that the ‘final 3’ TIFF images would just be ignored if the 4th image, to complete the set, was missing.
1 Like
I think I may have found my error; I included the pictdata.txt file in the selection. I’m trying again as I type, but excluding that file and so far the local processing seems to be working (294 x TIFF single layer images)
I am using the LaQuinta multispectral camera, flew at 30m altitude with 75% overlap.
1 Like
Spoke to soon, that dataset has also failed. Where can I get the processing log from? Is there somewhere I can access through the WebODM portal?
1 Like
Well my camera produces 5 bands per image but the same processing issue would occur. I have not tried running a data set in which I know of a missing image (just one of the bands). If these guys said that WebOdm would respond as a failed data set, then I would just accept that as a correct answer.
I would rather see the data set complete and report that an image was not used (a single image made up of all required bands). I know that in my case at least that I would almost certainly still have enough overlap to still produce a good result.
2 Likes
It just occurred to me that it may be interesting to test your WebODM with a different data set. If you would find that interesting, I would be happy to share a small data set for you to run. It would be one that I have processed successfully. This should isolate for you that your data set is the issue if this is at all a concern for you. Just let me know. There are 1810 images in this. It is the smallest I have.
Maybe this one.
2 Likes
Hi, that would be awesome, yes please! Thanks 
1 Like
Done. I will PM the details.
1 Like
Make sure there a no spaces in the file names and also, give it a try with a much smaller dataset where you know you have the correct number of band-images for each capture. ODM can fail due to both of those things.
-
Here’s a set of 205 multispec images (link) that I’ve verified can run on webODM, so give that a try and see if they work on your system (although these are dji multispec images so if the issues is caused by some weird file format issue this won’t help)
-
See also discussion here
-
This code snippet might be of use if you need a somewhat automated way to make sure you have all bands for each capture.
2 Likes
Hi Tim,
What special technique did you use to upload to webODM? I’ve downloaded the files (thank you very much for posting them!) and I assume the files I need to process are the .tiff contained in the ‘odm-texturing’ folder. However, I can’t select multiple folders to upload and this folder contains further folders split by band, eg Green, Red etc.
I thought I might create a single folder containing all the Tiffs from each of the sub-set band folders, but can’t do that either as the files have duplicate names (think that’s a DJI thing).
I think I might just have to pay for Pix2dFields as that seems to actually work with no problems. I did pay for webODM but it’s really not worth the hassle!
Hi, I still can’t figure this crappy software out. I’ve downloaded the files (thank you again!) and have two zipped folders containing the tiff files that I should process.
But of course I can only process one of the two folders, and amalgamating the two folders to one just causes file naming issues which still breaks webODM.
Am I doing something drastically wrong here? Doing the same thing and expecting a different result is the definition of madness!
webODM only expects the tiff files, right? Or does it need any of the other files?
Nothing to do with multispectral issue you’re having, but you mentioned multiple folder/upload issues. I had the same problem because DJI puts 1000 images in a dir and if there are more it creates more dirs with 1000 images in each dir. I found a mass file renaming tool (rename a bunch of files in a single go) for my OS that worked great. Once all files were renamed with unique names, I was able to merge them all into a single directory and upload that. Search for “rename multiple files for x” where x is your operating system. The files don’t all have to be sequentially named, but they do have to be unique.
2 Likes
Sorry, I was in a hurry when I posted that link but it wasn’t the correct one for the original files. This link should work to download the original (renamed) files. Name it something sensible (current default name is the unhelpful “images.zip” 
) and once you have unzipped it, delete the “console_output.txt” from the folder so there are only images in there.
Hopefully that should work.
Sorry this is all frustrating, you are not alone in that experience! I could fill a book with the number of hours I’ve spent having things fail on drone projects. And adding multispec adds another layer of misery on top of the usual drone related dramas. Unfortunately it seems to come with the territory, although we’re all trying to make it less painful as fast as possible!
I think Agisoft and pix4d do have easier to use workflows for multispec data so you may find that is worth the investment. Many of us are working on making it easier but it is still a work in progress! (and when companies like dji put out badly designed tools like the Phantom 4 multispec, it just makes everyone’s life harder)
Note though that with pix4d at least, once your data is in there the only way to get it back out is to download each of your projects one at a time, by hand. That’s probably fine if you only use it a bit, but if you’re like me and have 300 projects in there and don’t have time to download them all, one at a time, before the 6-month window after you stop paying for their license, they will happily delete all of your data forever.
1 Like