vvvv » Contributions
» VVVV.Packs.MachineLearning (alpha)
This site relies heavily on Javascript. You should enable it if you want the full experience. Learn more.

VVVV.Packs.MachineLearning (alpha)

pack
Credits: motzi, César Souza, devvvvs, woei

about

This is a first release of the MachineLearning pack, bringing several methods of the field to VVVV, mostly by wrapping functionality from the great accord.net framework by César Souza.

This is a first release which is still lacking most of the documentation and has not been heavily tested.

There is some girlpower though.
Many of the features will be explained in the upcoming workshop at NODE15.

More detailed info to come...

Features

  • Naive Bayes Classifier
  • k-Nearest Neighbour Classifier
  • Kernel Support Vector Machines
  • Hidden Markov Models (and HMM classifiers)
  • k-Means clustering
  • Principal Component Analysis
  • Kernel Discriminant Analysis
  • Deep Belief Networks (experimental)
  • ...

Upcoming

tekcor 24/04/2015 - 19:25

seriously amazing!
What was your motivation to implement it? pure interest or something specific?

edit: and which vvvv bersion do you recommend for it?

motzi 25/04/2015 - 11:33

most stuff i developed with alpha33.8 and did some tests with b33.7, which did not show any problems. earlier versions might work a well but are untested.

my motivation is a melange of personal interest in the field and some ideas i've been carrying around in my head for quite some time :)

499reality 11/05/2015 - 12:32

really interesting.. thanks for this hard work

guest 15/05/2015 - 00:46

Really good pack Motzi. Can you recommend somewhere I could go to understand the what is funamentally going on here so I could understand the application of my own datasets?

The many of us who were not able to make your Workshop would probably like to understand what is going on a little better. Not that there is anything wrong with the GP patches, they're laid out very clearly, but novices like me need some dumb love.

everyoneishappy 15/05/2015 - 04:15

Looks amazing! Also really curious to know more. Of all the (many) workshops missed at NODE, I think this is one of the most regret inspiring for me :)

motzi 15/05/2015 - 11:52

sorry for being sloppy on updating this - unfortunately lots of work had to be done right after node that kept me from doing this right away.

slides of the workshop and few more examples will follow asap.

Steryl 02/06/2015 - 14:56

which patch did you use for the facial recognition example that you made in the workshop? I would like to try some things with it.

guest 08/07/2015 - 18:10

More examples would be really great!
Btw, the QueueStore turns red (does not exist in the current beta). Copied from the alpha addons, but it's not working bith beta 64.

isdzaurov 03/05/2016 - 10:08

Hi and thank u motzi !

But why QueueStore node red?

joreg 03/06/2016 - 11:57

hei @isdzaurov better use @motzi to get his attention..

motzi 09/06/2016 - 13:08

@guest and @isdzaurov: hmm, i didn't notice this yet. i'll have a look at it soon...

motzi 09/06/2016 - 15:05

@isdzaurov: i cannot reproduce the QueueStore turning red. can you give an example patch (maybe it's better to open a forum thread)

isdzaurov 21/02/2017 - 11:10

Hi motzi

You will not prompt where example "face expression recognition example" ?

thanks

id144 21/02/2017 - 12:22

@isdzaurov it is in x86 version
vvvv_50beta35_x86\packs\VVVV.Packs.MachineLearning\girlpower\face_recognition.v4p

It is because FreeFrame FaceTracker lib used by the example is x86 only. I can imagine it's quite easy to replace it with Kinect HDFace and get much more compelling results.

michaelsed 29/04/2018 - 22:35

Hi @motzi

Thank you for this very great contribution :)

Is there a chance that you will do a new workshop on how to make use of the Hiden Markov Model Trainer and the others present in the pack as you said by the end of this workshop:
https://www.youtube.com/watch?v=XtYRMjJcrwA&t=6558s

motzi 04/05/2018 - 17:28

hi @michaelsed,

thanks for your kind words.
to be honest i did not get the success i was hoping for when doing body gesture recognition but there is room for further research. so at the moment i do not have a good data set to demonstrate how they could be applied successfully.

what do you have in mind to use it on?