Factors influencing detection of CCTag3 markers

[Eric Chameroy]

Hi everyone,

I've noticed that my CCTag3 markers seem to always be detected by FeatureExtraction, but not always detected by FeatureMatching and/or StructureFromMotion. What would be the cause of this? Is this something I should be concerned about, especially if I'm hoping to use CCTag3 for scaling?

Thanks for the help.

Best,

Eric

[Eric Chameroy]

I've attached this to provide some clarity to what I'm talking about:

Notice that at the bottom, CCTag3 is 3/0/0. I did have CCTag3 checked for all three modules.

[Fabien Castan]

Hi Eric,

As the CCTag encodes a unique ID, you could avoid the geometric validation on the FeatureMatching.

You can use a different FeatureMatching node to perform the matching of the CCTags on which you could disable the geometric validation.

And then you can connect the output of the 2 FeatureMatching nodes (one with CCTag and the other one with SIFT&AKAZE) on the input of the StructureFromMotion node.

Best,

Fabien

--
You received this message because you are subscribed to the Google Groups "AliceVision" group.
To unsubscribe from this group and stop receiving emails from it, send an email to alicevision...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/alicevision/0c4a2f8d-a755-491f-8e57-23c684628a4fn%40googlegroups.com.

[Eric Chameroy]

Hi Fabien,

Thank you for your help regarding my questions.

I'm confused about how to hook up the FeatureMatching2 node. Should the connections look similar to those of FeatureMatching? Should I run connections from 'input' and ''featuresFolders' to those in StructureFromMotion? I tried, but nothing happens. Also, when you say geometric validation, are you referring to 'Geometric Filter Type'?

Best,

Eric

[Eric Chameroy]

Hi Fabien,

I think I figured out the pipeline. Does this look correct?

Too bad they're not color coded like in the manual example. Do some of these just pass through like 'featureFolders'?

Here are some Display Feature results:

Sorry, I still don't understand what is happening here. I thought the sets of numbers referred to the number of features detected during FeatureExtraction, FeatureMatching, and StructureFromMotion because in those modules I have to select the feature types. I don't know why there's still a zero for cctag3.

Best,

Eric

[Fabien Castan]

> Does this look correct?

yes

> I don't know why there's still a zero for cctag3.

The viewer only display the feature/matches/landmarks for the last selected node.

So, as we have used 2 nodes for the FeatureMatching, you see the matching results only for one of them (the one with sift and no cctag).

---

De : alice...@googlegroups.com <alice...@googlegroups.com> de la part de Eric Chameroy <eric.c...@gmail.com>
Envoyé : lundi 9 novembre 2020 21:53
À : AliceVision <alice...@googlegroups.com>
Objet : Re: Factors influencing detection of CCTag3 markers

 

⚠️ Do not click or open unknown attachments ⚠️ **

To view this discussion on the web visit https://groups.google.com/d/msgid/alicevision/2bdbd1e2-9dee-424c-ba49-b0821ef2d591n%40googlegroups.com.

[Eric Chameroy]

Thanks Fabien!

Eric