Cluster
| Author: | chmee |
| Date: | 19 Jan, 2011 |
| Category: | plugin |
Download
64bit
v1.0 - complete recoding
v0.8 - inital contributed version
Description
Cluster is a simple algorithm for merging Pointclouds. May be useful for simplifying contourdata. sometimes it will be reasonable to use two Clusters in row..
ONLY >=v25
- v0.8 - simple Algorithm coded
All hints and experiences are greatly appreciated. discussionthread - http://vvvv.org/forum/following-tracked-objects-(n%3E1)
v1.0 Changelog
based on dbscan, now it works far more robust against noise. changed category from value to 2d. some extras like lifecycle, motionvectors & automatic linear movement after losing tracking, affinity of the points (0=noise, 1=empty, then the clusters) and so on, try it and leave your messages, ideas & hints here or in the thread linked above.
Comments
Comments are no longer accepted.Please create a new topic in the vvvv beta forum to discuss this contribution.
@chmee
Really Nice Contribution! It could be useful for lot of things..
ah well, the helpfile is referencing to Chaos.v4p which is a Subpatch of the Group (2d) module i dropped somewhere in the 2.0-tikiwiki-forums.
so everybody starting this helppatch with a "virgin"-b25 installation will have 2 red nodes called "Chaos"... unless this subpatch can be found in
dear chmee, pls don't take the following as offense: dear devvvvs, pls DO take the following as offense:
after having a closer look at **Cluster (Value) help.v4p** using the free XMLmarker (must-have!) i encountered a funny thing again:
vvvv's xml makes use of (at least two) new attributes called systemname and filename. again those new attributes are neither documented nor part of the(https://vvvv.org/glossary) of vvvv
have to complain about this in the bug section.
no problem. good to find this way :D v1.0 is on the way and the help-patch will be a native one.. (didnt thought about your chaos ;)
it's only partly my chaos :D
of course you can use "my" chaos! but you should distribute it then....
He Chmee, thanks for this great contribution. Is there a way to avoid multiple ID allocation?
missed that one looks great :) tx!
Unless I misunderstand what the output pins are, "Members" is listing the cluster number that each X & Y input slice belongs to, correct?
But I'm seeing numbers in the "Members" output pin that are not in the "ID" output pin list. There's not even a very good overlap; for example, "Members" may contain 0, 2, 3, 4, 5, but "ID" only contains 4 slices of 0, 1, 2, 3. Note this is with constant input data so nothing is changing with time.
How do these pins correspond to each other? What does the "Members" data mean? I had hoped to be able to use that pin to group corresponding Z data, but no joy. Thanks!
I just noticed that in the help patch that the dots and the corresponding boxes usually do not have the same color, indicating mismatched IDs.
And I also just saw a case where the ID list had a number twice (6, 1, 5, 6) - should that be possible?
-edit-
OK, so I'm guessing that in Members the IDs are +2 to make room for the affinity values 0 and 1, but the double 6 in the IDs still puzzles me.