Regression in DTM in latest build

I recently did an ultra/ultra run on a dataset and got a result that significantly differs from previous data, at least as seen from the report.

This is what I expect, or at least there’s no real surprises. The rectangular angry bit in the north central section is a pond and still water causes wackiness.

There’s now this big crater where the pond and the trees were, which did not appear the last time I ran this task. :frowning:

What else can I do to help figure out what’s going on here?


It’s the middle of the night for me so forgive me if I’m way off, but it looks like you are comparing the DSM and the DTM, and them being different from one another makes sense.

Do you have them both from prior versions? Or the data to share?


They should be different, but not in this way. For instance, the house and barn are neatly erased from the DTM, but the huge crater is definitely wrong.

Here’s an example (different data but same location) that was processed on April 5th:

This is what I recall seeing all the time previously. I’m going to save all the products from this run and then rerun it with a freshly downloaded container and see what I get. I do not know precisely which settings I used on the original one, so I’ll try ultra/ultra first since that’s what I used for the original post.

1 Like

Hmm :thinking: Do you think you tweaked the smrf parameters? If so, replicating them could be difficult without them being documented.

1 Like

I have yet to examine that level of fine-tuning – feature and PC quality are enough for me at the moment. :slight_smile:

The cratering is even worse here!

The images total about 1.5GB and have been uploaded to Hub now.

1 Like

Images resized to 2000 pixels to save time.


188 images 14:06:33
(14 hours? No, actual time from QR Processing Time 1.0h:41.0m:51.0s)

|Processing Node:|node-odm-1 (auto)|
|Options:|auto-boundary: true, dem-resolution: 1, dsm: true, dtm: true, feature-quality: ultra, gps-accuracy: 8, mesh-size: 250000, orthophoto-resolution: 0.5, pc-filter: 5, pc-geometric: true, pc-quality: high, pc-rectify: true, resize-to: -1, use-3dmesh: true|
|Average GSD:|5.09 cm|
|Area:|61,558.93 m²|
|Reconstructed Points:|21,925,492|


First, thank you so much for running the dataset and sharing the output.

Second, now I am confused as to why we are seeing different results!

I will try again with settings like yours and see what I get.


I ran another several missions today and then processed the images with my arguments to see if there was something about those pictures, and I’m still seeing the cratering. I’ll do the same with Gordon’s.


Indeed, I am surprised you’re having bad reconstruction in that area. What does your report’s overlap analysis look like? This was a clean dataset. Great work!

Created on:	6/26/2022, 1:50:08 PM
Processing Node:	node-odm-1 (manual)
Options:	auto-boundary: true, debug: true, feature-quality: ultra, feature-type: orb, min-num-features: 15000, pc-quality: ultra, skip-3dmodel: true, skip-orthophoto: true, verbose: true, rerun-from: dataset
Average GSD:	2.47 cm
Area:	61,492.06 m²
Reconstructed Points:	132,200,975
1 Like

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.