Re: [help-GIFT] gift features

[Mika Rummukainen]

Hi,

thanks to both of you for the information.

Second question I had in mind was about the order of the numbering of the
features - I mean what feature labels or numbers correspond to which features. I
just wanted to verify I got the order right.
For me, it seems that the features are numbered by the following order:

1. color blocks

- 1st 256 times 16x16 color blocks, where each has 166 possible values 
  (i.e. features 0..42 495, where 0..165 are from the first block etc.)

- 2nd 64  32x32 blocks,
  (10 624 more feats, i.e. 42 496 .. 53 119)

- 3rd 16  64x64 blocks,
  (2 656 more, i.e. 53 120 .. 55 775)

- 4th 4  128x128 blocks
  (664 more, i.e. 55 776 .. 56 439)

2. color histogram
166 features
 (feats 56 440 .. 56 606)

3. gabor blocks
- 256 times 16x16 blocks, each block containing 120 possible Gabor features
 (features 56 607 .. 87 325 so that in 56 607..56 726 are the Gabors calculated
from the first block and in 56 726..56 845 are from the second etc.) 

4. gabor histogram
last 120 features ( 87 326 .. 87445 )


cheers,
Mika

> Hi,
> 
> > If you do not omit the lowest frequency band of the Gabor filters
> (which
> > might be the default in gift) you end up with 30720 Gabor blocks and
> 120
> > features for the histogram which gives a total of 87446 features.
> 
> Line 495 from gift-add-collection.pre-pl tells me that we write out 
> InvertedFileFeatureDescription.db using the gift-write-feature-desc
> command 
> whose source you can find in FeatureExtraction/write_feature_descs.c .
> This command should use the exact same settings as are used for the
> feature 
> extraction. If I am not mistaken the original author of that file (as
> of the 
> feature extraction code) is David Squire. Is there anything to add (or
> to 
> change in the code), David?
> 
> > I hope that this makes things clearer and I hope that I calculated
> > correctly.
> 
> Looks like it :-)
> 
> Cheers,
> Wolfgang
>