Can intermedia be trained to do the image comparison

Question

I already sent email to Larry, I think he is working on my questions now, and I tried to use the functions to compare two teeth image signatures which from one person,same teeth pictures with some difference, then I got totally different results.
float lscore1
= OrdImageSignature.evaluateScore
( img_sig1, img_sig2, "color=0 texture=0.9 shape=0.1 location=0" );
System.out.println("case1 color=0,texture dominate>>>" + lscore1);
//>>>11.3160
float lscore2
= OrdImageSignature.evaluateScore
( img_sig1, img_sig2, "color=0 texture=0.1 shape=0.9 location=0" );
System.out.println("case2 color=0,shape dominate>>>" + lscore2);
//>>>>87.6091
float lscore3
     = OrdImageSignature.evaluateScore
     ( img_sig1, img_sig2, "color=0 texture=0.5 shape=0.5 location=0" );
System.out.println("case3 color=0,weigh half half>>>" + lscore3);
//>>>49.463
float lscore4
     = OrdImageSignature.evaluateScore
     ( img_sig1, img_sig2, "color=0 texture=0.2 shape=0.7 location=0.1" );
System.out.println("case3 color=0,shape dominate, texture major, location minor>>>" + lscore4);
//>>>79.6232
since intermedia user JAI internally, I am wondering if the comparison method used by JAI or intermedia are based on neural network knowledge, if so, is there a way to train intermedia for specific application like mine--teeth image comparison? can the comparison only focus on the shape of one or two teeth in the middle of the pictures?
thanks for any answer,
Tianjian Wang

Answer 1

No, we are not based on neural network technology, and currently there is no way to train interMedia for specific domains. Also we currently do not have the ability to focus on one or two teeth in the middle of the picture (that is, focus on a region).