Which of the two processing methods will give me the most accurate results? I’ve used both methods on the same stock pile and got two different results. The results show a larger volume on the cloud processing vs the desktop and it is usually within 5-8% difference. I know I should stick with one method to maintain accuracy but now that I can process faster through the cloud I want to opt for that method as opposed to the desktop method. Any thoughts?
in general, accuracy of results should not depend on the processing platform and we expect similar results regardless if you are processing on the Pix4Dcloud or in Pix4Dmapper.
Would it be possible that you share both quality reports (Pix4Dcloud and Pix4Dmapper)? I would like to have a look if there was something wrong with the processing that could affect the final results.
How do I send the reports to you?
There is an option to attach it directly to the reply here on the community (upload button). Else you can upload them here.
How do I send you the desktop results? I did both methods and got ~4% difference between the two.
Thank you for sharing the Quality reports. I had a look and the processing looks almost identical on both Pix4Dcloud and Pix4Dmapper.
As for the differences in volumes, how do you ensure that the vertices of volume boundary are exactly in the same spot on both Pix4Dcloud and Pix4Dmapper volume?
If they are not exactly the same then this could explain the difference you get for volume measurements.
P.s.: I removed the quality reports so they are not publicly available anymore.
When it comes to ensuring the vertices are in the same location I merely do a visual comparison between the two projects. I basically, follow the same track using the same start point for both projects. You’re right, if it is not exact then I will get different volumes but a 4% error seems rather large for very minor deviations on well a defined stockpile. The other factor that I’m taking into consideration is that in **“every”** project comparison I’ve done the cloud version is always a higher volume. So despite minor discrepancies in outlining a stockpile I would assume that I would get variations in the volumes going one way or the other but %100 of the time ( and I’ve done several) it favors the cloud. Seeing as how I plan to continue to do these projects for some time I think I will try and incorporate some GCP’s next time. Maybe the errors are coming from a deviation in the initial project elevation.
It is possible that there are differences between the volumes calculated in Pix4Dmapper or Pix4Dcloud due to:
- The project was processed on separate platforms and it can happen that results are not identical.
- The volume calculation on the Pix4Dcloud was optimized for online computations and has a slightly different approach than the Pix4Dmapper one.
- To be able to compare the two volumes, exact same points should be used (not the case in the project you shared).
It is hard to say which calculation is more accurate as we do not have the ground-truth of the stockpile.
My expectation is that when using exactly the same point for the base of the volume, the final calculated values should be similar and within the range of the estimated volume error.
Please let us know if you notice big differences in the volumes also in the future.