Skydio Datasets

Thinking about buying a Skydio and processing all images through odm → Does anyone have experience with running skydio datasets through ODM? I realise skydio is randomly flying around structures, thus adding lots of images with sky included.
Your feedback / experiences so far would be interesting to discuss.

1 Like

Nothing yet, though I agree, it’d be wonderful to get hands-on with Skydio data.

If you get one, would you be willing to help seed some test data for us?

Here you go - a telco tower. Can you access/download?
https://drive.google.com/drive/folders/1xKBe1Eb9rVDLaEoNOOexc_BFZERrAt9I?usp=sharing

80% overlap. 300+ images

I am particularly interested in how we might deal with corrupted reconstruction when sky shows in images. Totally aware of odm thriving with mapping and terrain/environmental modelling, but maybe there is a way to get structures reconstructed well with this amount of imagery.

This is what I get with skydio software.
image

1 Like

Use the --auto-boundary option when processing a dataset; it will filter out anything outside of the camera location boundaries (including the sky). You need the latest version of ODM for that option.

1 Like

Thanks for sharing a Skydio dataset! I don’t think anyone had shared one yet, so we’ve never been able to test it officially.

2 Likes

I just finished a run with that vs no auto-boundary. I hope to have screenshots soon (provided the no auto-boundary actually finishes, which it may not).

Created on: 10/26/2021, 8:18:40 PM
Processing Node: Lightning - UAV4Geo (manual)
Options: auto-boundary: true, cog: true, crop: 0, debug: true, dem-gapfill-steps: 4, dem-resolution: 1, dsm: true, dtm: true, matcher-neighbors: 16, mesh-size: 300000, min-num-features: 16000, orthophoto-resolution: 1, pc-classify: true, pc-ept: true, pc-filter: 1, pc-geometric: true, pc-quality: high, use-3dmesh: true, verbose: true
Average GSD: 0.09 cm
Area: 145.69 m²
Reconstructed Points: 14,957

image

image

hblanken-9234-autoboundary_Report.pdf (2.7 MB)


I don’t see the Skydio in the OpenSFM database, so I’m not sure if adding it might help reconstruction any.

I’ve opened a PR:

1 Like

Thank you @saijin_naib - this is consistent with my results. Surprising to me.
Good idea reg. OpenSFM.

I can get further datasets if this helps?

1 Like

Yes, if it isn’t too much trouble for you!

I’d love to see other collection types like more typical aerial field survey, maybe a building or wall survey…

But whatever you can provide is excellent.

This alone was helpful, and this is a super cool dataset. More example datasets will help even more.

ezgif-3-689c9bbe1cde

I made some good progress, but won’t be able to work on this more until the New Year. But maybe we should work on putting together a funding campaign for improvements for orbited objects like this at that time, with special emphasis on quality of cell tower outputs.

2 Likes

Amazing! Looks almost better vs the skydio render :wink:

Let me get some more datasets of ‘orbited’ objects.

2 Likes

@smathermather @Saijin_Naib

Do you like a challenge :slight_smile:
Try this:
https://drive.google.com/drive/folders/1Q5o98WQm9bMBCgbeL3_TDmhmY0viR4H_?usp=sharing

2 Likes

Stephen’s doing some magic to his stuff, so I’m quite excited to see what he can tease out of that dataset!

I’m a bit more reserved, but happy with the current state of what I can pull from just loading/processing the data as-is:

Created on: 10/28/2021, 4:05:47 PM
Processing Node: Lightning - UAV4Geo (manual)
Options: auto-boundary: true, cog: true, crop: 0, debug: true, dem-gapfill-steps: 4, dem-resolution: 1, dsm: true, dtm: true, feature-type: orb, matcher-neighbors: 16, mesh-size: 300000, min-num-features: 160000, orthophoto-resolution: 1, pc-classify: true, pc-ept: true, pc-filter: 0, pc-geometric: true, pc-quality: high, resize-to: -1, use-3dmesh: true, verbose: true
Average GSD: 0.19 cm
Area: 154.54 m²
Reconstructed Points: 155,449

image

I hope I can get something reasonable out of the pylons, too!

Created on: 11/1/2021, 1:48:00 PM
Processing Node: Lightning - UAV4Geo (manual)
Options: auto-boundary: true, cog: true, crop: 0, debug: true, dem-gapfill-steps: 4, dem-resolution: 1, dsm: true, dtm: true, feature-type: orb, matcher-neighbors: 0, mesh-size: 300000, min-num-features: 1600000, orthophoto-resolution: 1, pc-classify: true, pc-ept: true, pc-filter: 0, pc-geometric: true, pc-quality: high, resize-to: -1, use-3dmesh: true, verbose: true
Average GSD: 11.62 cm
Area: 708.75 m²
Reconstructed Points: 9,576,421

image


Created on: 11/1/2021, 3:49:31 PM
Processing Node: Lightning - UAV4Geo (manual)
Options: auto-boundary: true, boundary: {"type":"FeatureCollection","name":"hblanken-9234-12-boundary","crs":{"type":"name","properties":{"name":"urn:ogc:def:crs:OGC:1.3:CRS84"}},"features":[{"type":"Feature","properties":{},"geometry":{"type":"Polygon","coordinates":[[[150.793827317975,-34.435710430832785],[150.79383151001545,-34.43555099346689],[150.7941398630476,-34.43555655868901],[150.79413567159193,-34.435715996087914],[150.793827317975,-34.435710430832785]]]}}]}, cog: true, crop: 0, debug: true, dem-gapfill-steps: 4, dem-resolution: 1, dsm: true, dtm: true, feature-type: orb, matcher-neighbors: 0, mesh-size: 300000, min-num-features: 1600000, orthophoto-resolution: 1, pc-classify: true, pc-ept: true, pc-filter: 0, pc-geometric: true, pc-quality: high, resize-to: -1, use-3dmesh: true, verbose: true
Average GSD: 11.63 cm
Area: 707.82 m²
Reconstructed Points: 809,994

image

Marginally better, but I think at this point we need smathermagic.

This is a nightmare. Absolutely achievable with tuning, but my word… .

1 Like

Also, as someone who has stayed the heck away from 330kv installations, this is painful to look at. Wow. I love that this is possible to fly.

2 Likes