Viola-Jones IMMI Extension
panagiotious
New Altair Community Member
Hello,
I am trying to set up a face recognition project on Rapid Miner using the IMMI image processing extension, but I seem to be kind of stuck! I have set up a Viola-Jones operator to get me a model and a Detector to find the faces in an image. My training set consists of 100 face images (very small resolution) and 100 urban and nature photos (medium resolution). Since my set was too small, I was trying to find some cascades to get it going. Found a site for openCV (http://alereimondo.no-ip.org/OpenCV/34), wrote a java XML parser to convert the cascades, but found out a lot more values than the 3 immi has. Some "trees", 5 values per feature per stage and so on. Could anyone help me figure out which values I need for my implementation?
Moreover, after getting the POIs from the detector, I cannot find any way to crop the image and use the faces in order to feature-extract and then use a classifier for the final classification process... Is there some way that can happen? I was told (by IMMI) that there is no operator to get the coordinates of the segments and parse a new image to a feature extractor. Is there something I am missing?
Thank you for your time,
Panagiotis Kintis.
I am trying to set up a face recognition project on Rapid Miner using the IMMI image processing extension, but I seem to be kind of stuck! I have set up a Viola-Jones operator to get me a model and a Detector to find the faces in an image. My training set consists of 100 face images (very small resolution) and 100 urban and nature photos (medium resolution). Since my set was too small, I was trying to find some cascades to get it going. Found a site for openCV (http://alereimondo.no-ip.org/OpenCV/34), wrote a java XML parser to convert the cascades, but found out a lot more values than the 3 immi has. Some "trees", 5 values per feature per stage and so on. Could anyone help me figure out which values I need for my implementation?
Moreover, after getting the POIs from the detector, I cannot find any way to crop the image and use the faces in order to feature-extract and then use a classifier for the final classification process... Is there some way that can happen? I was told (by IMMI) that there is no operator to get the coordinates of the segments and parse a new image to a feature extractor. Is there something I am missing?
Thank you for your time,
Panagiotis Kintis.
Tagged:
0