Re: [help-GIFT] more patches

[risc]

Now that cvs access is figured out...

I have applied the patch to extract_features.c, so that current CVS compiles, 
and functions.

In line with patch policy, I'll commit my next set (attached to the origional 
mail) come wednesday, unless instructed otherwise.

This patch was just to get CVS functional. The next patch brings my test from 
1.82 to 1.45 or so, so major improvement.

Julia Longtin <address@hidden>

On Thu, Aug 03, 2006 at 09:31:51AM -0500, address@hidden wrote:
> Well, first, i'd have to know what you're talking about. ;P
> 
> I'm not familiar with any methods of using ssh keys for authenticating to a 
> CVS
> server. i'm familiar with CVS over SSH, but i always used a password with 
> mine.
> 
> mind pointing me to a document on the version of cvs/ssh you're using?
> 
> other than that, sounds good. i can abide by 'commit if broken, otherwise
> post to list for a few days, then commit if no-one complains'. ;)
> 
> i added changelog entries this time, they look alright to you?
> 
> Thanks again,
> 
> Julia Longtin <address@hidden>
> 
> 
> On Thu, Aug 03, 2006 at 08:34:34AM +0200, Wolfgang Mueller wrote:
> > Dear address@hidden (sais my muttrc)
> > Dear Julia,
> > 
> > Up till now, due to historical reasons, I liked to keep an eye on the
> > CVS and apply patches myself. I will not dive into project history here,
> > but I would propose you send me your ssh key for inclusion into the list
> > of those who can access the gift CVS. 
> > 
> > However, I would find it good if bigger changes continue to be discussed
> > in CVS.
> > 
> > Do you like that idea?
> > 
> > Cheers,
> > Wolfgang
> > 
> > On Wed, 02 Aug 2006, address@hidden wrote:
> > 
> > > well, first off, you seem to have dropped a patch during the last set.
> > > gift compiles, but dosent work.
> > > 
> > > i'm attatching that patch to this mail. please apply immediately.
> > > 
> > > second off, i'm adding another patch, that colapses kernel[11,12,21,22] 
> > > into one variable, kernelsxy, and removes it from the global scope.
> > > 
> > > speed improvement: 1.72 VS 1.66. probably just because of less calls of 
> > > malloc().
> > > 
> > > Thanks,
> > > 
> > > Julia Longtin <address@hidden>
> > 
> > > --- ../../dev2/gift/FeatureExtraction/extract_features.c  2006-06-28 
> > > 00:08:54.000000000 +0000
> > > +++ FeatureExtraction/extract_features.c  2006-06-28 21:50:05.000000000 
> > > +0000
> > > @@ -22,15 +22,35 @@
> > >  #define numV 3
> > >  #define numGrey 4
> > >  
> > > +#define OLDFIXED
> > > +#ifdef OLDFIXED
> > > +/* the number of scales to recognise features at */
> > > +#define num_gabor_scales 3
> > > +/* the number of color blocks we're going to break the image into 
> > > (overlapping) */
> > > +#define num_total_colour_blocks (256+64+16+4)
> > > +#endif
> > > +
> > > +void init_feature_variables(uint32_t, uint32_t **, uint32_t ***, 
> > > uint32_t ***);
> > > +void extract_gabor_features(PPM *, uint32_t ***, uint32_t ***);
> > > +void extract_mode_features(PPM *, uint32_t *, uint32_t, uint32_t **, 
> > > byte *, uint32_t *);
> > > +enum ppm_error write_mode_features(char *, uint32_t, uint32_t *, 
> > > uint32_t ***, uint32_t ***, byte *);
> > > +
> > >  int main(int argc, char *argv[]) {
> > >  
> > >   char *in_fname, *out_fname;
> > >   char *point_pos;
> > >   FILE *ppm_file;
> > >   PPM *im_rgb, *im_hsv, *im_quant;
> > > - int *colmap, colmap_size;
> > > + uint32_t *colmap, colmap_size;
> > >   enum file_types ppm_type;
> > >   enum ppm_error the_error;
> > > + uint32_t ** block_gabor_class[num_gabor_scales];
> > > + uint32_t ** gabor_histogram[num_gabor_scales];
> > > + /* the "mode" of each color block. */
> > > + byte block_mode[num_total_colour_blocks];
> > > + /* the colour count of each block. for quantizing the image */
> > > + uint32_t * col_counts[num_total_colour_blocks];
> > > + uint32_t * col_histogram;
> > >   
> > >   switch(argc) {
> > >   case 2:
> > > @@ -94,17 +114,20 @@
> > >           exit(1);
> > >   }
> > >  
> > > + /* FIXME: move this back into init_feature_variables. */
> > > + col_histogram = (uint32_t *)malloc(colmap_size*sizeof(uint32_t));
> > > +
> > >   /* initialise the variables required for the feature extraction */
> > > - init_feature_variables(colmap_size);
> > > + init_feature_variables(colmap_size, col_counts, block_gabor_class, 
> > > gabor_histogram);
> > >  
> > >   /* extract the features */
> > >  
> > > - extract_mode_features(im_quant, colmap, colmap_size);
> > > + extract_mode_features(im_quant, colmap, colmap_size, col_counts, 
> > > block_mode, col_histogram);
> > >  
> > > - extract_gabor_features(im_hsv);
> > > + extract_gabor_features(im_hsv, block_gabor_class, gabor_histogram);
> > >  
> > >   /* write them to the file */
> > > - if ((the_error = write_mode_features(out_fname, colmap_size)) != 
> > > PPM_OK) {
> > > + if ((the_error = write_mode_features(out_fname, colmap_size, 
> > > col_histogram, block_gabor_class, gabor_histogram, block_mode)) != 
> > > PPM_OK) {
> > >     ppm_handle_error(the_error);
> > >     exit(1);
> > >   }
> > 
> > 
> > > _______________________________________________
> > > help-GIFT mailing list
> > > address@hidden
> > > http://lists.gnu.org/mailman/listinfo/help-gift
> > --
> > Dr. Wolfgang Mueller
> > LS Medieninformatik
> > Universitaet Bamberg
> 
> 
> _______________________________________________
> help-GIFT mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/help-gift