Impact of DJI Smart Oblique Imagery on PIX4Dmatic Processing

PIX4Dmatic PIX4Dmatic Questions/Troubleshooting
1

We’ve conducted tests comparing datasets captured with the DJI Mavic 3E RTK using Smart Oblique and standard oblique modes. Each dataset comprised approximately 1,500 images of a construction site. Notably, processing the Smart Oblique dataset in PIX4Dmatic required at least five times longer than the standard oblique dataset.

We are aware that oblique imagery can increase processing times due to the complexity and volume of data. However, the substantial difference observed raises concerns about potential inefficiencies or issues specific to Smart Oblique imagery.

Could anyone provide insights into why Smart Oblique imagery might cause such a significant increase in processing time? Did we do something wrong? Are there recommended settings or workflows to optimize the processing of Smart Oblique datasets in PIX4Dmatic? Smart oblique reduces the flight time dramatically, and we wish to use this mode if possible.

2

I am also experiencing long process times as well as wrong camera calibration. I suspect it is because the orientation angles (yaw,pitch, roll) are those from the drone and not the gimbal, I am still waiting for user inputs and a response for support…

3

Months are already gone without even an answer. Hello support??? Whats the workflow here?

4

Sorry for the delay in answering. Indeed, this does not seem to be handled well today and we need to look into that. Are all Smart Oblique flights problematic or does it happen in specific cases for you? Would you be willing to share some problematic datasets with us, maybe even with the analysis you’ve made? This could speed up the process to find a solution.

@drones6 feel free to create a ticket in case there is an urgent request, you can click on Pixie

image
image100×107 5.2 KB
and from there it’s possible to directly reach out to a support agent.

5

@Pierangelo_Rothenbuhler - I just messaged you a link to a smart oblique dataset for this site - District Line Rd - Chandler - NIRA

This was done with the DJI M3E. I have other datasets available from the M3E, as well as the M300+P1 if needed.

Thanks

1 Like
6

Thanks @Derrick_Westoby1, shared with the team for further investigation

1 Like
7

Again, any news here?

8

Hello,

Thank you for your patience.

We are currently investigating this issue and are planning to include performance improvements for the calibration step in one of the upcoming releases.

In the meantime, please let us know if you encounter similar performance issues in other processing steps with DJI Smart Oblique projects.

Kind regards,
Alexey

9

Hi!

Has a fix to this been implementet? What is the estimated time we get a beta or stable release?

1 Like
10

Would be nice to get a status on this issue! :slight_smile:

1 Like
11

Hello there, 4 months gone already, no news??

12

Hello,

Thank you for your patience. Our R&D team has completed a thorough analysis of the datasets that we received. I’d like to share our findings with you.

We conducted two rounds of testing. Our initial tests did not reveal any significant difference in calibration performance. To ensure a comprehensive analysis, recently we processed the datasets through the entire workflow, from calibration to mesh generation, using our latest stable release, version 1.81.

The results of this recent analysis were consistent with our initial findings. We observed only marginal performance differences, with the smart oblique dataset calibrating slightly faster than the nadir datasets.

Based on these tests, we have been unable to reproduce the specific performance degradation for DJI Smart Oblique datasets.

To help us investigate further, it would be extremely helpful if you could provide another dataset where you have observed this performance issue.

We are committed to resolving this for you. Please let us know if you are able to provide an additional dataset or if you have any other questions.

Best regards,

Alexey