Re: [help-GIFT] Features Extraction

[Wolfgang Mueller]

Dear Epa Uwimana,

On Tue, 28 Nov 2006, Epa Uwimana wrote:

> Here is the script:
> 
> open FILE,shift;
> read(FILE,$lString,4);
> print unpack("I",$lString),"\n";
> while(read(FILE,$lString,8)){
>       print join(":",unpack("If",$lString)),"\n";
> }
> I figured out that I had a typo in my script but even after fixing it, I
> am getting a lot of number that I have no ideas what they are and now I
> am wondering if they are right.

You did not read the code, right? To understand what the script is doing
I suggest reading the perlfunc manpage.

First: this script works only on 32bit machines.
This script reads a 4byte integer that is coded as a binary. Then it
reads features while the file is still readable (i.e. not everything has
been read). After reading 8 bytes, it will turn them into an
integer/float pair and output both numbers separated by a ":".

The integer is the feature ID, the float is the term frequency in the IR
sense.

> Try to run the attached image and let me know what the result would mean
> I am interested in knowing what every number stands for i.e RGB,HSV,
> Energy,Histogram and others

This information you get by reading the first few papers about Viper
that you can get at viper.unige.ch. The mapping from number to feature
you get by running

gift-write-feature-descs < /tmp/gift_tmp31287.ppm                              

where /tmp/gift_tmp31287.ppm stands for a 256x256 ppm image. There
should be one gift_tmpYYYYY.ppm image in your tmp ;-) after indexing images.

Cheers,
--
Dr. Wolfgang Mueller
LS Medieninformatik
Universitaet Bamberg