Final Imagery Processing Failing or stuck in infinite loop?

[SColchester]

This project: https://drone.hotosm.org/projects/molino-de-yera

Selecting 'Start Final Processing' does not seem to result in the processing starting. Can select option again and again.

[linear]

TECH-1162

[spwoodcock]

At least this provides a better error message now, but it's not fixed.

I think the tasks need to be in READY state, instead of HAS_IMAGERY. Will fix so this is possible with multiple states on the individual tasks.

[SColchester]

Nice, yes because I understand that users may often want to jump to creating a large ortho immediately and skipping the task by task processing

[spwoodcock]

We have actually been considering two possible workflows:

1. What we have now - task processing is a validation step - but at the end we submit all photos and process (requiring lots of resources).
2. An alternative where each task processing produces quality orthos, then the final 'project level' step is actually just stitching all the parts together using processing tools, reducing the overall resources required (not processing all the images at once, instead in small batches).

The second approach seems more resource efficient & makes the most of the data 'splits' we are artificially introducing by virtue of using task areas to collect imagery - I like this idea, if it works well! (without seams / artifacts at the joins).

But from a user experience perspective, do you have a preference on what would be best best here @SColchester? 🙏
(i.e. processing each task, then one more step to generate the ortho. or a single 'process all imagery' button to do it in one go)

[SColchester]

@spwoodcock thanks for summarising, from my perspective my preference would be: a single 'process all imagery' button to do it in one go.

Comparing the results of (1) doing several smaller webODM runs (i.e. task sized) and then stitching the results together vs (2) doing large webODM runs, I get much better orthos from #2 - typically with far fewer visible seams. For me the big value add of this cloud processing via DroneTM is that I might be able to big imagery runs which would otherwise overwhelm my system if I were to do it offline with webODM.