search

samsoe / mpg_aerial_survey

0 stars 1 forks source link

230612_spurgepoly job 5 has issues #27

Open mosscoder opened 1 year ago

mosscoder commented 1 year ago

https://storage.googleapis.com/mpg-aerial-survey/surveys/230612_spurgepoly/post_process/odm/report_5.pdf

Screenshot 2023-06-14 at 11 32 36 PM

The photos included in this job are spread all over the place.

I'm going to rerun the job, hoping for the best. In parallel I'll scrutinize if the photo list is filtering properly for this job. Weird that it would work on 230601 but not this one if it failed at the filtering step.

mosscoder commented 1 year ago
Screenshot 2023-06-14 at 11 43 58 PM Screenshot 2023-06-14 at 11 43 38 PM
mosscoder commented 1 year ago

I don't see anything crazy about the photos assigned to this job. I think the photo filtering step before ODM is fine:

Screenshot 2023-06-14 at 11 51 27 PM Screenshot 2023-06-14 at 11 51 49 PM
mosscoder commented 1 year ago

GCPs assigned to it (red) look fine:

Screenshot 2023-06-15 at 12 11 55 AM
mosscoder commented 1 year ago

Filtered GCP list

Filtered manifest

Action items: -Run without GCPs -Run without --force-gps -Inspect photos in the manifest. Anything visually evident? Water on the lens?

samsoe commented 1 year ago

It is also curious over 60 images were rejected. 1101 over 1164 (94.6%)

mosscoder commented 1 year ago

Lead number 2:

https://community.opendronemap.org/t/odm-fails-where-pix4d-succeeds/7840/8

I suspect if we restrict the default quality control measures, we might get a little better result. Worth testing. This would require a dynamic method to pass options to ODM.

mosscoder commented 1 year ago

@samsoe can you dig into the logs and retrieve any errors suggesting a given photo is rejected?