The Lamborghini

You clicked because of the title, didn’t you? Well fear not car fans, because I have for you a genuine survey, of a genuine Lamborghini, made from genuine yellow… plastic. Click bait? You-betcha.

So my son has this crappy toy Lamborghini that I took some freehand photos of using my camera phone with GPS turned off. I did this because I’ve been doing hundreds of surveys of various things and turning all the knobs on ODM to get a better understanding of what does what. Since I started, I’ve picked up the ODM book which I’d recommend - it would have saved me a lot of time if I had it when I started.

I couldn’t find a dataset that I could use for testing out the 3D modelling options in the way that I wanted, which was being able to add freehand photos to a regular model to improve the detail in some areas. All of the photos are freehand at various angles and pitches.

Dataset: https://dashboard.aerosurvey.co.nz/files/shared/222.tar.gz
Size: 377 MB
Images: 73

Firstly, I know a lot of the options below contain Ortho / DEM options. I have some presets that I use for surveys for consistency.

So here’s the first try from a month or two ago. This one came out remarkably well; the best of the lot in fact, (which is the reason for this post):
Lambo (Original)
https://dashboard.aerosurvey.co.nz/public.cgi?S=222&T=8b0e533f-88c1-4b19-82f8-4dd38cc81c38
Options: fast-orthophoto: false, use-fixed-camera-params: true, resize-to: -1, dsm: true, dtm: true, crop: 10, use-3dmesh: true, depthmap-resolution: 800, dem-resolution: 2.0, orthophoto-resolution: 2.0

It’s not perfect, but under the circumstances I think the results are great. I left it at that for a while and focused on other survey tests, but I came back to the Lambo data a couple of weeks ago as I noticed that there were new presents in WebODM that I was keen to run the data through. I’d say that the new surveys are, uhhh, “different”. A muddy puddle? An overcooked egg? Modern art? You decide.

Lambo (3D Model)
https://dashboard.aerosurvey.co.nz/public.cgi?S=272&T=d032b931-b885-4403-af9a-29869624524d
Options: fast-orthophoto: false, use-hybrid-bundle-adjustment: false, use-fixed-camera-params: true, resize-to: -1, dsm: false, dtm: false, crop: 10, use-3dmesh: true, dem-resolution: 2, orthophoto-resolution: 2, depthmap-resolution: 1000, opensfm-depthmap-min-patch-sd: 1, min-num-features: 8000, matcher-neighbors: 8, texturing-data-term: gmi, use-opensfm-dense: false, mesh-size: 600000, mesh-octree-depth: 11, texturing-nadir-weight: 16

Lambo (POI)
https://dashboard.aerosurvey.co.nz/public.cgi?S=273&T=cb9f31ec-2530-414e-9fe5-c772782a49db
Options: fast-orthophoto: false, use-hybrid-bundle-adjustment: false, use-fixed-camera-params: true, resize-to: 2048, dsm: false, dtm: false, crop: 10, use-3dmesh: true, dem-resolution: 2, orthophoto-resolution: 2, depthmap-resolution: 800, opensfm-depthmap-min-patch-sd: 1, min-num-features: 8000, matcher-neighbors: 24, texturing-data-term: gmi, use-opensfm-dense: false, mesh-size: 600000, mesh-octree-depth: 9, texturing-nadir-weight: 16

Lambo (Forest)
https://dashboard.aerosurvey.co.nz/public.cgi?S=286&T=421f893c-81af-4485-8b34-4ec9076da16a
Options: fast-orthophoto: false, use-hybrid-bundle-adjustment: false, use-fixed-camera-params: true, resize-to: 2048, dsm: false, dtm: false, crop: 10, use-3dmesh: false, dem-resolution: 5, orthophoto-resolution: 5, depthmap-resolution: 1000, opensfm-depthmap-min-patch-sd: 1, min-num-features: 18000, matcher-neighbors: 21, texturing-data-term: area, use-opensfm-dense: false, mesh-size: 100000, mesh-octree-depth: 9, texturing-nadir-weight: 16

Lambo (Buildings)
https://dashboard.aerosurvey.co.nz/public.cgi?S=287&T=3caaba6d-50d1-4ffb-ab01-d764f31d3d9d
Options: fast-orthophoto: false, use-hybrid-bundle-adjustment: false, use-fixed-camera-params: true, resize-to: 2048, dsm: false, dtm: false, crop: 10, use-3dmesh: true, dem-resolution: 2, orthophoto-resolution: 2, depthmap-resolution: 1000, opensfm-depthmap-min-patch-sd: 1, min-num-features: 8000, matcher-neighbors: 8, texturing-data-term: gmi, use-opensfm-dense: false, mesh-size: 300000, mesh-octree-depth: 10, texturing-nadir-weight: 28, rerun-from: odm_orthophoto

Lambo (SFM) - This one is actually good, though noisy.
https://dashboard.aerosurvey.co.nz/public.cgi?S=305&T=3e2a3c24-fa64-49e6-9074-be9c342bd358
Options: fast-orthophoto: false, use-hybrid-bundle-adjustment: false, use-fixed-camera-params: true, resize-to: -1, dsm: true, dtm: true, crop: 10, use-3dmesh: true, dem-resolution: 2, orthophoto-resolution: 2, depthmap-resolution: 800, opensfm-depthmap-min-patch-sd: 1, min-num-features: 8000, matcher-neighbors: 8, texturing-data-term: gmi, use-opensfm-dense: true, mesh-size: 100000, mesh-octree-depth: 9, texturing-nadir-weight: 16

Lambo (TDT-A) - Texturing Data Term: Area; This one has done a fantastic job at recreating the fence… and the egg.
https://dashboard.aerosurvey.co.nz/public.cgi?S=306&T=18251052-f827-4742-a484-4e2fa2016d29
Options: fast-orthophoto: false, use-hybrid-bundle-adjustment: false, use-fixed-camera-params: false, resize-to: -1, dsm: true, dtm: true, crop: 10, use-3dmesh: true, dem-resolution: 2, orthophoto-resolution: 2, depthmap-resolution: 800, opensfm-depthmap-min-patch-sd: 1, min-num-features: 8000, matcher-neighbors: 8, texturing-data-term: area, use-opensfm-dense: false, mesh-size: 100000, mesh-octree-depth: 9, texturing-nadir-weight: 16

Lambo (Volume)
https://dashboard.aerosurvey.co.nz/public.cgi?S=309&T=62d3da8d-d748-4ac4-9ec4-e8e697a06e5c
Options: fast-orthophoto: false, use-hybrid-bundle-adjustment: false, use-fixed-camera-params: true, resize-to: -1, dsm: true, dtm: false, crop: 10, use-3dmesh: false, dem-resolution: 2, orthophoto-resolution: 2, depthmap-resolution: 1000, opensfm-depthmap-min-patch-sd: 0, min-num-features: 8000, matcher-neighbors: 8, texturing-data-term: gmi, use-opensfm-dense: true, mesh-size: 100000, mesh-octree-depth: 9, texturing-nadir-weight: 16

Lambo (Hi-Res) - The same options as the original survey…
https://dashboard.aerosurvey.co.nz/public.cgi?S=285&T=17367f76-0ece-409b-9bcf-94e5719e620a
Options: resize-to: -1, use-3dmesh: true, orthophoto-resolution: 2.0, dtm: true, dem-resolution: 2.0, crop: 10, dsm: true, depthmap-resolution: 800, use-fixed-camera-params: true

Original options and it’s… still an egg? Somebody’s having a yoke and my eggspence.

I’ll have a fiddle with the options on the good ones and see how good I can get it.

3 Likes

Yes, I clicked on that because the titIe promissed that you deal with a smaller objekt, not a landscape. I just started today with WebOM, I have some experience with VisualSFM and I’am looking for a programm to get 3D Modells from smaller parts. Your notice was the only text I found that gives me some help to play with the parameters. Unfortunately I cannot download any link, they all are expiered. Is there any way to get at least the dataset?
Best regards Norbert

The links say they are not completely processed.

ODM is a fast moving project and these results were processed about half a year ago. Around early May 2019, an update changed the processing of 3D models and the quality wasn’t the same - there seemed to be a lot more noise and inaccuracy in the output. So I set out to test many different settings to attempt to recreate a good model. I never managed to create a model that was as good as the model from pre-May 2019, but I shared my results anyway to help others to get good results from models with the new version.

The model outputs were relevant at the time, but were processed 6 months ago, so I removed them recently as ODM has been updated many time since then and this thread had also died. I’ve re-run 4 of them for you with the above settings matching the description so that you can see the results from today.

Default: https://dashboard.aerosurvey.co.nz/public.cgi?S=429&T=f53f879c-4e96-4c32-802f-8d9e9ab53382
3D Model: https://dashboard.aerosurvey.co.nz/public.cgi?S=428&T=79504d6b-c531-46ed-9101-c6f8541e3b04
Forest: https://dashboard.aerosurvey.co.nz/public.cgi?S=430&T=7580c8e2-5dfb-41b6-a063-11d928d88fe7
POI: https://dashboard.aerosurvey.co.nz/public.cgi?S=431&T=5311efd3-98e2-49c3-8319-5600893fc454

The dataset is available in the link at the top - feel free to process further and share your results.