[vlfeat] annotated tag debian/0.9.17+dfsg0-2 created (now a28788b)

Dima Kogan dima at secretsauce.net
Sun Dec 22 21:05:37 UTC 2013


This is an automated email from the git hooks/post-receive script.

dkogan-guest pushed a change to annotated tag debian/0.9.17+dfsg0-2
in repository vlfeat.

        at  a28788b   (tag)
   tagging  fa5775a9ea649d058a2050eb049cb93ba3a32ebe (commit)
  replaces  v0.9.6
 tagged by  Dima Kogan
        on  Sun Dec 22 13:04:32 2013 -0800

- Log -----------------------------------------------------------------
vlfeat Debian release 0.9.17+dfsg0-2

Andrea Vedaldi (834):
      toolbox: avoids creating the shortcut version.m to vl_version.m as this conflicts with MATLAB's version command
      Makefile: adds missing libtool flag for Mac OS X 10.5 compatibilty
      Makefile: removes false errrors caused by missing *.d files
      Makefile: simplifications
      mser and sift: warnings and spaces
      test_stringop.c: fixes warnings
      mathop.h: adds a function to convert a vector comparison type to a symbolic name
      mathop.c: comments
      moves preprocessor macros from generic.h to host.h
      generic.h: adds a function to convert a data type identifier to a symbolic name
      host.h: completes definition of printf formatting stings for builtin datatypes
      mexutils.h: suppresses some spurious GCC warnings
      mexutils.h: adds IN() and OUT() macros to easily access input and output parameters of the MEX file
      mexutils.h: adds vlmxIsPlain() test
      mexutils.h: comments
      mexutils.h: adds two default error identifies
      vl_binsum: updates code to latest version of mexutils.h and resolves GCC warnings
      comments
      vl_getpid.c: fixes a subtle compatibility problem with Windows.h
      Makefile: improves reliability of distribution system
      0.9.6 -> 0.9.7
      doc: adds a note about 0.9.7
      Makefile.mak: makes selection of mexopts.bat automatically dependent on the architecture
      xtest: new Matlab toolbox unit tests
      integrates xtest in Makefile and vl_setup.m
      vl_setup.m: fixes a bug preventing the QUIET option from working correctly
      vl_twister.*: updates code to new style
      vl_imsmooth.*: rewrite to use new styles and to simplify the code
      eliminates obsolete imop.*
      imopv.*: simplifies the code
      generic.c: comments
      vl_sift.m: spaces and comments
      mexutils.h: reorder the code
      vl_dsift.c: makes messages more uniform with the rest of the library
      vl_test_binsum.m: fixes a case not tested correctly
      Makefile: adds missing VLDIR dependencies
      adds experimental support for multithreading
      new Visual C++ project
      rewrites heap implementation
      XCode: adds reference to index.html
      new vl_imdisttf.m to compute distance transform
      new shuffl-def.h: metaprogram to randomly permute a sequence
      new qsort-def.h: metaprogram implemeting quick sort
      heap-def.h: style
      test_heap-def.c: changes VL_PRINTF to printf
      new kmeans implementation (operates on floats and doubles rather than integers)
      Xcode: updates to latest changes
      adds VL_UNUSED attribute to supporess certain GCC warnings
      kdtree: adds support for FLOAT and DOUBLE data and cleans up the code
      refactors mexutils.h and removes obsolete functions
      Xcode: updates the project to the latest changes
      vl_test.m: clear functions
      Makefile: style
      Makefile.mak: updates to latest changes
      restores Windows compatibility
      generic.h: fixes a bug preventing correct compilation without threading support
      kdtree.h: fixes 64-bit compatibility
      vl_alldist: fixes a crash occurring when the second argument had null size
      vl_binsum.c: fixes 64-bit compatibility
      Makefile: turns on more compiler warnings
      random.*: makes helper functions inline and adds functions to generate random indexes in a range
      qsort-def.h: makes indexes unsigned
      shuffle-def.h: removes 32 bit indexes limitation
      test_qsrot-def.c: resolves a GCC warning
      kmeans.*: fixes 64-bit compatibility and cleans-up
      stringop.h: fixes 64-bit compatibility and updates style
      random.c: resolves a warning
      mser.c: resolves some warnings
      src/sift.c: resolves some warnings.
      vl_alldist.c: updates style
      generic-driver.h: resolves some warnings
      src/check.h: basic support for unit testing
      test_stringop.c: makes it a unit test
      Makefile: breaks it down to submodules
      fixes a symbol missing on some platforms (true -> VL_TRUE)
      Makefile: bugfixes
      test_*: fixes compiler warnings and 64 bit issues
      mathop.c: fixes 64 bit issue
      random.c: fixes compiler warning
      vl_simdctrl.c: fixes argument checking
      vl_inthist.c: fixes 64 bit issues
      vl_getpid.c: fixes compiler warnings and argument checking
      vl_binsearch.c: fixes 64 bit issues, updates style, and cleans up
      vl_imsmooth.c: fixes compiler warning
      vl_imdisttf.c: fixes 64 bit issue
      vl_imdisttf.c: mxuError -> vlmxError
      toolbox/**/*.c: muxError -> vlmxError
      vl_sift.m: fixes 64 bit issue
      vl_hikmeanspush.c: fixes 64 bit issue
      vl_kmeans.c: fixes datatype
      kdtree.h: fixes 64 bit issue
      samplinthist.tc: fixes 64 bit issue
      vl_alldist2.c: fixes 64 bit issues
      vl_inthist.c: improves style
      vl_localmax.m: improves comments
      vl_ubcmatch.c: fixes compiler warning
      vl_kdtreequery.c: fixes typo
      xtest: adds vl_assert_exception to check for exceptions
      xtest: vl_test_twister.m: adds a check for invalid arguments
      xtest: vl_test.m: adds the possibility of running a single suite
      vl_twister.c: improves style and fixes 64 bit issues
      mexutils.h: changes the prefix of exceptions to vl:
      vl_getpid.c: restores Windows compatibility
      update-mak.sh: adds a check to see if the Makefile.mak needs update
      update-mak.sh: adds AUTORIGHTS
      Makefile: imrpoves style
      vl_alldist.c: improves argument checking
      vl_binsearch.c: improves argument checking
      kmeans.*: completes implementation of vl_kmeans_init_by_copy
      vl_kmeans.c: adds an option to repeat k-kmeans multiple times
      vl_imdisttf.c: improves argument checking and fixes special cases in which parabolas are flat
      vl_binsearch.m: adds a unit test and improves comments
      vl_whistc.m: replaces MEX version with vl_binsearch / vl_bimsum instead
      kmeans.c: improves comments
      Xcode: updates to latest changes
      Makefile: fixes a64 target
      vl_imdisttf.c: fixes a reference to an obsolete function
      imopv.*: adds vl_imintegral_*
      vl_imintegral.*: rewrites to use library function
      doxygen.conf: fixes spaces and discards VL_UNUSED
      mexutils.h: fixes comment
      random.h: fixes comment
      stringop.c: fixes the name of a variable
      stringop.c: fixes the name of a variable
      vl_imsmooth.c: fixes undefined symbol crash
      mathop.h: improves comments
      generic.h: adds a function to return the size of a data type
      generic.h*: adds a function to return the current CPU time as a double
      kmeans: simplifies interface moving to library logic for multiple iterations
      moves update-mak.sh to make/
      random.*: improves comments
      kmeans.*: fixes typos
      vl_figaspect.m: imrpoves style
      new vl_test_kmeans.m unit test
      Xcode: fixes ARCHS build setting for proper debugging on 64 bit target
      new vl_colsubset.m function
      new homkermap
      mexutils.h: cosmetic
      vl_test_kmeans.m: comment
      vl_imdisttf.c: fixes memory leak
      vl_imwbackard: fixes argument checking
      xtest: improves style
      generic.h: fixes a warning
      finishes implementation of homkermap
      mathop, vl_alldist: adds support for Jensen-Shannon kernel.
      homkermap: adds support for Jensen-Shannon kernel.
      Makefile and code: makes VL_ENALBE_* and VL_*_ENABLED consistnet
      Makefile: now debugging must be eplicitly activated by declaring the DEBUG symobl. This symbol triggers the VL_DEBUG preprocessor macro.
      improves functions to report the library status in human readable form
      Makefile: minor cleanup
      Makefile: fixes a bug for which noprefix was not created
      vl_imsmooth.c: fixes a bug causing images with more than three channels to be handled incorrectly. Fixes the style of the error messages.
      Makefile: fixes some Octave icompatibilities
      adds workarounds to improve compatibility with Octave MEX interface
      Makefile: adds a PROFILE switch to turn on debugging symbols with optimization
      fixes incorect VL_SUPPORT_SSE2 to the correct value VL_ENABLE_SSE2
      moves the image distance transform code from the MEX file to the library
      float.th: fixes a bug that was preventing the correct VSIZE to be set for DOUBLE
      vl_setup.m: generates an error message when an incorrect option is passed in
      Xcode: activates PROFILE option for release target
      rodrigues.c: fixes a special case
      mathop.c: fixes a bug that caused the Jensen-Shannon disance function to return an undefined value
      Uses abort() rather than assert(0) to mark unreachable default cases in switch statements. Contrary to assert(0), abort() correctly marks an exit point also when debugging is switched off, preventing the compiler from generating spurious warnings.
      sift.c: fixes an incorrect variable initaliziation
      improves kd-tree demos and adds them to the default demo suite
      vl_demo_print: uses vl_printsize and adds an option to specify which figure to print
      vl_errfill.c: fixes a signed comparison warning
      .gitignore: adds directory opt/
      vl_printsize.m: adds more options
      vl_demo_print.m: fixes a bug introduced with the latest changes
      vl_imdisttf.c: fixes an incorrect error message
      vl_setup.m: makes greetings less verbose
      kdtree.c: comments
      image distance transform: fixes a sign convention
      Makefile: simplifies enabling/disabling of MATLAB and Octave support
      Makefile: new help target prints instructions
      Makefile: changes style of titles printed by the info target
      adds a toolbox/mex subdirectory adding MEX files one level up in the hierarchy
      Makefile.mak: syncrhonize with latest changes
      restores Windows compatibility
      starts the kdtree tutorial
      adds missing .html
      test_threads.c: fixes a warning
      adds image distance transform tutorial
      image distance transform: fixes a bug preventing the transform to be computed with inf values in the data
      doc: improves matlab installation instructions
      doc: improves compliling instructions
      doc: improves fron tpage and look
      generic.h: 0.9.7 -> 0.9.8
      vl_argparse: fixes convention to pass in a cell array of options rather than a comma separated list
      mdoc: changes the way description lists are rendered
      improves comments
      adds DSIFT tutorial
      dsift.c: improves comments
      vl_dsift.c: fixes error messages.
      improves comments
      updates COPYRIGHT messages
      adds vl_xmkdir.m
      adds vl_phow.m
      updates README
      Makefile.mak: 0.9.7 -> 0.9.8
      Makefile: fixes a typo
      Makefile: restores compatibility when the shell is not bash
      make/dist.mak: fixes bugs and improves code style
      vl_imbackwardmx.c: fixes a bug arising in some special cases
      adds pegasos
      pegasos: various refinements
      pegasos: significantly improve speed and adds the ability of continuing the computation
      kmeans.h: comments
      vl_pegasos.c: fixes a corrputed messages
      host.{h,c}: replaces code violating C99 aliasing rules
      webdoc.py: adds support for including text verbatim
      adds caltech-101 example application
      web.css: tweaks style of boxes
      adds missing file vl_pegasos.m
      Makefile.mak: adds pegasos
      fixes pegasos in windows
      vl_homkermap.c: configures MATLAB environment correctly
      vl_argparse.m: comments
      vl_override.m: comments
      vl_printsize.m: uses a fixed page size as reference and resize the page to enclose the figure tightly
      vl_kmeans.c: fixes a printed message
      README: adds more details on v0.9.8
      Merge remote branch 'origin/master'
      vl_phow.m: changes a default parameter value
      vl_phow.m: adjust tuning and completes the options
      phow_caltech101.m: makes the demo much faster with a only slight decrease in performance
      phow_caltech101.m: moves result in a independent model object
      vl_phow.m: comments
      phow_caltech101.m: adds kdtrees
      phow_caltech101.m: comments and restores tiny problem as default
      adds apps/sift_mosaic.m
      sift_mosaic.m: works without arguments with default images
      data/a.jpg: makes if of the same size of data/b.jpg
      adds test images
      0.9.8 -> 0.9.9
      phow_caltech101.m: comments
      README: adds v0.9.9
      sift_mosaic.m: simplifies test run
      Makefile: moves a comment
      Makefile: fixes a case in which version.html was not updated
      doc: adds references to v0.9.9
      adds an image for the apps page
      doc: typos
      vl_demo_kdtree_sift.m: improves detection of FLANN
      adds vl-config script
      vl_grad.m: comments
      vl_imarray*.m: refines comments and options, fixes style
      pegasos: add the ability of specifying a data permutation
      vl_imup.m, vl_imdown.m: updates the style
      vl_grad.m: update comments and handling of options
      generic.{h,c}: fixes incorrect initialization of the library state when threads are disabled
      new array.{h,c}
      mexutils.h: adds out of memory error
      mexutils.h: adds array.h support
      vl_argparse.m: comments and whitespaces
      VL_EXPORT: adds support for C++ compilers
      vl_kmeans.m: comments
      test_getopt_long.c: typo
      mexutils.h: adds a function to envelope a MATLAB array in a VlArray
      adds local binary patterns
      lbp.*: adds support for flipping the descriptors
      sift.c: fixes typo in comments
      vl_binsum.c: improves error messages
      matlab.mak: fixes bug passing an incorrect flag to Linux 64 bit target
      doxygen.conf: adds missing LaTeX packages
      adds several minor fixes to the API documentation
      getopt_long.*: fixes the documentation
      doxygen.conf: updates to Doxygen 1.7.2
      dll.mak: adds missing sourcing of the heard files
      generic.c: fixes the documentation
      host.*: minor documentation fixes
      stringop.*: minor documentation fixes
      dsift.*: minor documentation fixes
      vl_rcos.m: documentation fixes
      array.c: minor documentation fix
      Xcode: changes expected location of doxygen
      pegasos.*: adds preconditioner support and slows down initial iterations
      vl_test_init.m: recognizes one-lines test function definitions
      homkermap.*: adds gamma-variants of the kernels, extend them to the negative reals, and adds the support for choosing a windowing function. It also improves the speed and the documentation.
      vl_phow.m: adds FloatDescriptors option
      vl_argparse.m: minor documentation fixes
      vl_kdtreequery.c: corrects the name of an option to match the standard
      vl_roc, vl_pr: improves comments and style
      pegasos.c: changes the algorithm slightly and fixes computations with preconditioner
      vl_pegasos.c: fixes a possibly un-initialize variable
      vl_phow.*: adds support for more color spaces and adds basic test unit
      Xcode: fixes array.*
      Xcode: sorts files
      vl_grad.m: improves documentation and adds test unit
      vl_test_pegasos.m: fixes test unit settings
      vl_phow.m: improvements to the opponent SIFT version
      vl_xmkdir.m: removes obsolete function call
      vl_pr.m: adds an option to return stable precision and recall vectors
      vl_argparse.m: faster implementation
      vl_pr: optionally returns sorted scores
      vl_printsize.m: fixes a bug preventing the ASPECT option from working
      make: switches to use proper MATLAB architecture string
      adds Xcode 4 compatibility
      Makefile: Mac OS X default SDK to 10.6, deployment target still 10.5
      moves version to 0.9.10, adds a reference to the ACM award
      vl_ihashsum.m: adds characters to avoid mcc error
      fixes a number of Doxygen errors
      dist.mak: fixes automatic extraction of VER
      vl_roc.m: accepts slightly more general input format
      webdoc.py: adds support for syntax highlighting and cleans up
      doc: improves style, activates syntax highlighting
      improves some comments
      small documentation fixes
      homkermap.c: fixes compatibility issue with Visual C++
      homkermap.h: fixes the name of the guard
      make/udpate-mak.sh: fixes code to retrieve the version number
      Makefile.mak: updated
      compiling.html: adds more information
      matlab.mak: simplifies detection of MEX and makes MATLAB bin depend on that
      doc: minor improvements
      webdoc.py: fixes DocError()
      mexutils.h: implements missing vlmxWarning()
      vl_printsize.m: adds the MARGIN option
      kmeans: fixes typo in one of the symbol definitions
      kmeans.h: implements missing vl_kmeans_get_energy()
      vl: improves comments (no more Doxygen warninigs).
      Xcode: adds missing files, adjust options
      Makefile: makes maci64 the default architecture on the Mac
      Xcode: makes 2 spaces indentations the project defautl
      kmeans.h: fixes compilation issue
      matlab.mak: do not pack vl_tune_* functions in the noprefix set
      Makefile: bug fix
      Makefile.mak: fixes the spelling of the Windows architecture to match MATLAB convention (win32/win64)
      doc: fixes the spelling of the Windows architectures
      Makefile.mak: fixes a bug causing unwanted files to be inclued in the binary distribution
      make/dist.mak: changes default HOST for publishing
      generic.c: updates copyright notice
      apps/phow_caltech101.m: minor tweaks and sovles a compatibility issue
      mexutils.h: adds missing static keyword
      updates README to v0.9.10
      Makefile: fixes make help
      doc: fixes compilation instructions
      webdoc.py: fixes an issue with <br></br> tags not recongized by some browsers
      Makefile: fixes backward compatibility on Mac OS X
      Makefile: fixes potential bug
      make/bin.mak: faster compilation
      make/matlab.mak: fixes error message when MEX is not found
      make/matlab.mak: faster compilaton
      make/dist.mak: better comments
      vl_plotframe.m: fixes documentation
      Makefile: dump PROFILE on VERB
      vl/generic.c: fixes debugging messages
      vl_imarray.m: fixes a bug causing a window to be opened all the times
      toolbox:* fixes help messages formatting errors and whitespace issues
      Xcode: adds M-files
      make/doc.mak: fixes a bug for which VERSION was not updated correctly
      0.9.10 -> 0.9.11
      vl_compile: restores compatibility with latest VLFeat
      doc: fixes typo
      docsrc/install-matlab.html: fixes typo
      vl_compile.m: improves help
      fixes the linux 32 bit architecture string (glnx32 -> glnx86)
      v0.9.11 -> v0.9.12
      Makefile.mak: 0.9.12
      README: fixes arch naming
      doc: fixes a pointer to v0.9.12
      Xcode: refines a build scheme
      matlab.mak: copy rather than linking the dll (solves issues with tar instances that do not digest links)
      removes use of doc/version.html
      0.9.12 -> 0.9.13
      updates index.html to 0.9.13
      make/doc.mak: improves doc-deep target
      Makefile.mak: fixes a bug in the construction of the binary package
      doc: updates 0.9.13 comments
      Makefile.mak: fixes package creation
      make/dist.mak: improves reliability of the binary commits
      make/dist.mak: improves reliability, adds back doc
      vl_setup.m: generates a warning rather than an error when the MEX files are not found
      make/dist.mak: fixes bug causing doc to crash creation of package
      make/dist.mak: includes only what needed in doc
      changes two of the demo images
      make.doc: fixed rare dynamic linking bug with doc-deep target
      vl_setup.m: fixes issue when multipe VLeat libraries are installed
      Makefile: adds support for Mac OS X 10.7
      adds vl_alphanum.m and corresponding test function
      vl_imarraysc.m: fixes error message style
      matlab.mak: removes redundant code
      Xcode: upgrades MATLAB version
      kdtree.*: fixes the tuning of a parameter that may adversely affect the performance with low-dimensional data and more than one tree
      binsum: adds support for multiple storage classes
      Xcode: adds missing vl_alphanum.m
      generic.c: improves certain debugging messages
      new vl_plotstyle.m
      vl_colsubset.m: adds LARGEST and SMALLEST options; adds unit test
      vl_setup.m: be quiet by default
      updates the project for Xcode 4.2 and OS X 10.7.2
      vkmeans.c: fixes a bug causing the code to crash with NaN data
      vl_kmeans.c: avoids crash in corner case
      kmeans.c: adds restarting of empty clusters
      vl_printsize: improve comment
      vl_imarray*.m: improves handling of colour image arrays
      adds alpha SLIC superpixels support
      fixes an issue that caused a compilation error with stricter linkers
      fixes an issue that caused a compilation error with stricter linkers
      vl_setup.m: fixes issue when multipe VLeat libraries are installed
      Makefile: adds support for Mac OS X 10.7
      Xcode: upgrades MATLAB version
      updates the project for Xcode 4.2 and OS X 10.7.2
      doc: upgrade to Doxygen 1.7.5.1, adds a bibliography file, moves part of the documentation to use this
      slic: adds to Xcode, improves documentation, partial cleanup
      doc: fixes comments
      doc: updates doxygen css style
      webdoc.py: fixes bug in liking external css styles
      doc: fixes doxygen style
      kdtree.c: improves doc
      doxygen.conf: fixes @cite{} filter rule
      make: rearranges flags
      make: rearranges flags
      vlfeat.bib: adds new bibliographies
      {aim,drift,generic,homkermap,kdtree,kmeans,mser,pegasos,pgm,quickshift,sift}: updates doc and fixes whitespaces
      make: moves LIBTOOL elsewhere
      makefile: fixes message
      slic: updates doc and adds removal of small regions
      kdtree.c: fixes a memory leak (thanks to Philipp Fechteler)
      slic.m: fixes compilation error
      kdtree.c: fixes a memory leak (thanks to Philipp Fechteler)
      slic: adds documentations, clean-ups
      generic.h: 0.9.13 -> 0.9.14
      slic: fixes Visual C compatibility
      kemans.c: fixes Visual C compatibility
      Makefile.mak: tentatively extend the compatibility to MS Visual Studion 2008 and 2010
      mathop.h: adds math compatibility functions
      mathop.h: fixes
      mathop.h: fixes GNUC special cases
      generic.c: adds DEBUG info to version string
      slic.c: switches to internal rounding functions
      Makefile.mak: several fixes, support for DEBUG flags
      Visual C++: updates project
      doc: adds more 0.9.14 notes
      Makefile.mak: git -> $(GIT)
      Makefile.mak: bugfix
      generic.c: bugfix
      matlab.mak: allows setting MEX_FLAGS from command line
      make: enable correct recursion of make with CFLAGS, LDFLAGS, MEXFLAGS
      Makefile.mak: comments
      make: comments
      Makefile.mak: pop MSVCR files
      Makefile.mak: deeper distclean
      GPL -> BSD: changes headers of most files
      GPL -> BDS: updates COPYING
      GPL -> BSD: adds some missing notices
      GPL -> BSD: normalize newlines at the end, adds missing notices
      vl_sigmoid.m: comment
      vl_quickvis.m: adds notice
      GPL -> BSD: adds info to the README and the documentation
      removes copylet.txt
      GPL -> BDS: fixes more notices
      GPL -> BSD: fixes more notices and whitespaces
      GPL -> BSD: fixes notice in float.th
      normalise library headers
      doc: updates release information and news
      README: cosmetic fix
      Makefile: removes obsolete autorights and fixes the help target
      README: rewrite
      merges hotfix release
      COPYING: 2011 -> 2012
      Makefile.mak: fixes missing file in binary distribution
      make/doc.mak: fixes doc-deep and doc interactions
      doc: fixes copyright dates
      quickshift: doc fixes (thanks to David Doria)
      adds vl_det.m
      vl_roc.m: adds NumNegatives option
      toolbox: adds VL_CUMMAX()
      vl_roc.m and vl_pr.m: rewrites, more standard, less code, better doc
      mdoc.py: simplify identifier of MATLAB documentation pages
      web.css: better looking links within code sections
      web.css: define a new type of list which flows more naturally in text
      formatter.py: adds comments
      vl_pegasos.c: cosmetic fix
      doc: adds tutorial on ROC, DET, and PR curves
      host.*: improves documentation
      slic: fixes comments
      generic.c: adds developer documentation
      vl_cummax.m: comments
      mexutils.h: refactors a function name and adds string comparison functions
      apps/phow_caltech101.m: fixes call to vl_homkermap
      vl_kmeans.c: avoids crash when an empty matrix is passed in
      vl_test_assert_almost_equal.m: now works recursively on structures
      vl_pr.m, vl_roc.m: fixes a bug treating -inf scores
      Makefile.mak: fixes compilation under MSVC 2010
      vl_xmkdir.m: update style and improves code comments
      Xcode: adds support for 4.3 and more general Makefile code that should work with any version
      make: fixes compilation with Xcode 4.4
      Xcode: update configuration
      vl_printsize: fixes comment style
      vl_demo_print: fixes comment
      doc: adds favicon
      doc: switches to new Google code
      doc: improves about
      doc: fixes about
      Xcode: adds some missing files in doc
      new HOG descriptors implementation
      hog: adds an option to compute features from polar representation of gradients
      hog: fixes rendering code
      vl_hog: more verbose
      hog: adds support for polar field input
      hog: improves doc
      hog: minor fixes and a tutorial
      hog: add bilinear interpolation of the orientation bins and fixes several bugs
      hog: makes bilinear orientation assignments configurable
      hog: adds a few options, fixes a few bugs
      Xcode: adds hog files
      vl_plotframe: cosmetic fixes
      make: lighter dependencies on DLL
      doc.mak: adds support for Inkscape figures
      aib: improves doc, adds verbosity control
      toolbox: removes some obsolete tests
      ihashsum: completes doc, fixes bug, adds test
      inthist: fixes a function name, comments, adds test
      removes more obsolete tests
      adds more xtests for aib, ikmeans, hikmeans, maser
      finishes removing obsolete tests
      adds alldist2 xtest
      slic.*: cosmetic fixes
      pgm.h: improves style
      pgm.c: fixes warning about unsigned quantities
      mathop: adds SVD for 2x2 matrices
      getopt_long.c: fixes a typo in the doc
      mexutils.h: fixes some data types and removes dead code
      vl_slic.c: fixes compilation warnings with Windows target
      generic.h: 0.9.14 -> 0.9.15
      generic.c: assigns authorship to the VLFeat Team
      shorter README with more pointers to the online documentation
      renames stock images
      new vl_impattern.m to generate stock patterns
      demo: fixes MATLAB demos
      README: mentions demos in the read me
      fix spaces
      small mods
      make/doc.mak: prints DOXYGEN variable when invoked with info
      make/doc.mak: maes doc-api dependent on bibliography file
      README: adds SVM stuff
      about.html: updates Vealdi's bio
      doxygen.conf: updates to doxygen 1.8.1.1
      web.xml: -> VLFeat Team
      Makefile.mak: updated
      fixes Windows compatibility
      demos: fixes file names after changing stock image names
      vl_demo: adds svmpegasos demo
      doc: improves front page and about
      vl_setup.m: avoids adding a directory that does not exist
      doc: several fixes and completes bios
      Xcode: reorders some files alphabetically, removes dead files
      Xcode: sorts more file, remove dead folder
      README: minor fix
      dsift.c: fixes datatypes
      mser.c: fixed datatype and format strings
      sift.c: fixes format strings
      generic.*: fixes format strings and types
      homkermap.c: fixes types
      ikmeans_*: fixes types
      impov.* fixes types
      kdtree.c: fixes types
      kmeans.c: fixes types
      pgm.*: fixes types
      random.c: fixes some types
      sift.c: fixes types
      slic.c: fixes types
      Xcode: add libvl.dylib native target
      README: minor improvement
      svm: general cleanup
      Xcode: adds missing files, reorders
      mdoc.py: adds generation of MATLAB toc file
      doc.mak: adds generation of MATLAB builtin doc documentation
      vl_help.m: switches to a more informative message
      adds missing toolbox/info.xml
      doc.mak: bugfixes
      svm: fixes copyright notices
      README: improvements
      doc: minor credit improvements
      update-mak.sh: do not include example files
      distribution and doc: fixes after renaming toolbox-src to mdoc.build
      doc: cosmetic fix
      Merge remote-tracking branch 'origin/master'
      removes a.jpg and b.jpg since links do not work in Windows
      hog: tutorial: typos
      Makefile.mak: uses -largeArrayDims for WIN64 target
      Makefile: adds -lpthread and -lm to Linux linker flags
      Makefile.mak: bugfixes
      removes dead vl_imtest.m function (see vl_impattern.m instead)
      modc.py: fixes generation of links
      doc: improves a few toolbox docs
      vl_impattern.m: adds blobs pattern
      covdet alpha
      imopv.*: adds image smoothing and gradient
      Xcode: adds covdet and native targets
      mat hop.*: adds simple linear system solvers
      covdet: get patches out
      covdet: adds affine shape estimation, many bug fixes
      vl_impattern.m: new blobs1 pattern
      covdet2.m: removes spurious code
      adds test_gauss_elimination.c
      imop.*: fixes constancy of the arguments of a function
      mathop.c: fixes gaussian elimination
      Xcode: adds test_gaussian_elimination.c
      covdet: adds shape adaptation, harris, starts laplacian scale estimation
      imopv.c: fixes major but in gaussian smoothing
      covdet: adds detection of laplacian scales
      covdet: adds SIFT descriptors
      imopv.c: solves warning
      covdet: refactors scalespace
      covdet: adds the ability to pass in custom frames
      vl_impattern.m: adds 'blob' pattern
      vl_covdet.c: small bug fix
      vl_impattern.m: adds options
      covdet.*: compensates for sub-smoothing in affine adaptation
      covdet: adds non-maxima suppression
      covdet: good perforamnce of hessian laplace
      covdet.*: some cleanup
      covdet2 -> covdet
      covdet: refactoring and cleaning up
      covdet: more refactoring and bug fixes
      covdet: refactos, and fixes Laplacian scale estimation
      covdet: fixes Laplacian scale even more
      covdet: nails down laplacian scale estimation
      covdet.c: some cleanup
      covdet: bugfixes in affine adaptation
      covdet: remove debugging info
      covdet.c: tweaks defaults
      covdet.c: excellent performance with hessian affine (accurate smoothing also works well, but not better)
      vl_impattern.m: fixes doc and adds an option to a pattern
      covdet: fixes orientations
      covdet: fixes sourcing of frames
      vl_sift.c: fixes typo in message
      covdet: fixes SIFT descriptors
      plotss.m: renames it
      vl_plotss.m: adds documentation
      vl_covdet.m: adds documentation
      covdet: bugfix and exports laplacian and orientation scores
      doc: adds covdet tutorial
      covdet: adds demo
      adds covdet tutorial to the homepage
      0.9.15 -> 0.9.16
      covdet: completes tutorial
      web.css: make site somewhat larger.
      covdet: integration of the tutorial
      Makefile.mak: 0.9.16
      covdet: fixes for windows
      scalespace: removes dead code, disable full doc temporarily
      covdet.c: minor cosmetic fixes
      index.html: adds VLBenchmakrs reference
      roadmap.html: updated
      vl_covdet.m: doc
      covdet: completes tutorial
      doc.mak: cosmetic
      vl_demo_covdet.m: shows how to use feature scores
      covdet.c: bug fixes (was missing half of the detections)
      vl_covdet.m: improves doc
      covdet: fixes typos
      covdet: adds more control and better non-max suppressions
      covdet: some more finetuining. TODO : allow to choose a threshold for Laplacian scale selection
      covdet.c: cosmetic fix
      about.html: updates Fulkerson's bio
      Makefile: leaves vlbenchmarks alone
      vl_demo_covdet.m: removes keyboard command
      sift: fixes the parsing of one of the options
      vl_demo_sift_cmd.m: style improvement
      lbp.h: adds missing guards
      vl_pr: adds normalised AP
      vl_assert_almost_equal.m: cosmetic fix
      vl_pr.m: fixes corner case
      scalespace: refactors and improves doc
      vl_demo_dsift.m: fixes bug causing figures to get mixed up in documentation
      webdoc.py: fixes bug with <script> tags
      doc: updates doxygen css form new base
      doc: refreshes style
      doc: switches to use MathJax for formulas
      doc: refreshes home page and about
      doc: minor fixes
      doc: more layout adjustments
      doc: improves layout on ie
      vl_covdet.m: fixes typo in doc
      adds vl_plotbox.m function and improves vl_linespec2prop implementation
      generic.*: includes support for OpenMP, hides better implementation details
      adds GMMs and Mahalanobis distance computation
      rodrigues.c: doc
      hog.c: cosmetic fixes
      pegasos.c: fixes compilation error with debugging info on
      doc: rearranges index in the front page
      doc: front page minor fixes
      doc: maser: adjust page
      doc: adds MathJax macros
      doxygen.css: adjust a few margins and dimensions
      web.xml: adds MathJAX macros
      kmeans: significant work on documentation
      kmeans: doc: minor fix
      vl_test*: cosmetic fixes
      .gitignore: cosmetic fixes
      vl_printsize: rewrites the function and adds a unit test
      Xcode: fully reorganise the project
      homkermap: several improvements:
      Random number generator:
      vl_argparse.m: raise a better error message when an option name is not a string
      vl_test_slic.m: quiet
      Major SVM fixes, alpha version.
      generic.*: rearranges pages, adds some developer documentation
      gmm: first cleanup pass:
      adds vl_threads: a MATLAB function to control VLFeat threads
      svm updates
      mathop.h: adds isnan and isinf predicates
      svm: adds support for all the losses, other minor fixes
      vl_svmtrain.c: fixes missing kernel in homkermap
      homkermap.c: major bug fix affecting expansion of features when the destination memory buffer was not cleared
      Makefile: creates two vesrsion of the library to address OpenMP instability.
      adds vl_frame2oell.m
      vl_plotsiftdescriptor.m: add support for affine frames
      kdtree.h: improves format of serialized KDTree
      vl_plotframe.m: improves doc and plot upright vector for rotation.
      starts fixing OpenMP support interface
      gmm: various improvements to the doc, refactors a few names
      minor OpenMP fixes
      host.*: restructure documentation
      kmeans.c: improves documentation
      cosmetic fixes
      cosmetic fixes
      Xcode reorganization
      make: fixes compilation on linux
      vl_svmtrain: fixes compiler warnings
      improves Windows nmake autodetection feature
      Makefile.mak: updates and fixes
      Merge branch 'vedaldi' of vlfeat.org:vlfeat into vedaldi
      Makefile.mak: bugfix
      test_liop.c: compiles under windows
      kdtree.c: compiles under Windwos
      liop: compiles under Windows
      mathop.h: compiles under Windows
      svmdataset.c: fixes compiler warning
      Visual Studio: updates project
      Makefile.mak: adds a note on OpenMP
      adds 0.9.17 release notes
      generic.c: starts restructuring the main documentation index page
      fix compilation under Linux 32
      Makefile: adjust linux flags to require GLIBC_2.3.4 only
      fixes compilation under matlab 09b
      fixes kmeans demos
      fixes compilation on mac 32 bit
      vl_demo_gmm_*: some improvements
      fisher*: reworks OpenMP
      generic.c: more doc
      vl_fisher.c: also fixes OpenMP
      gmm: refresh OpenMP
      web: updates style
      doc: improves layout
      doc: fisher vector
      gmm: properly controls number of threads
      Xcode: project fixes
      mexutils.h: fixes doc
      host.*: adds support to create integer constant literals portably
      random.*: improves portability
      kdtree.*: minor refactoring
      fisher.c: improves OpenMP implementation style
      vl_test_twister.m: minor fix
      kmeans.c: fixes OpenMP implementation style
      kdtree.*: improves interface, style
      svm: adds an option to weight the data non-uniformly
      generic.*: fixes compiler warnings
      mathop.h: adds sort
      vlfeat.bib: small fix
      vlad: improves doc, implementation, add options, add test
      Xcode: minor fixes
      test_gmm.c: fix to make it compatible with new VLAD
      fisher.c: doc
      Merge remote-tracking branch 'origin/milan_fixes' into vedaldi
      vlad: fixes Windows compatibility
      Makefile.mak: debug MEX files too
      Makefile.mak: fixes debugging options
      gmm.*: fixes OpenMP bugs
      Visual C: compiles using debugging symbols and no optimization
      mathop.h: fixes Windows support
      mathop.h: fixes
      Makefile.mak: fixes permission
      vl_gmm.c: improves style
      vl_demo_gmm_2d_twist.m: fix
      make dist: reduces clutter and improves speed
      dist.mak: fix
      doc.mak: fix
      vl_disft.*: enalbes choosing the geometry of the descriptor
      mexutils.h: bufix
      vlad: avoid NaNs on empty components
      vlad an fisher: refactor, fixes OpenMP bugs in fisher, adds test
      dist.mak: simpler and faster post-doc, make VER customisable from the environment
      Merge remote-tracking branch 'origin/vedaldi'
      Xcode: updates project to latest changes
      doc: reworks documentation system:  faster, easier to understand
      doc: various fixes
      doc: various improvements
      COPYING: new copyright message
      doc: more improvements, automatically generate toc in website pages
      fixes compiler warnings
      doc: further extensive improvement, better internal linking of documentation
      simplifies hana's loop, still numerically equivalent
      liop: simplifies how circle is computed . Almost near equivalence to old code (four pixels are different)
      loip: more simplifications, near equivalence maintained
      liop: major cleanup
      Xcode: sync
      vl_covdet.m: improves doc
      liop: further doc improvements
      removes obsolet vl_maketrainingset
      improves octave support and docs
      kmeans.c: bugfix
      phow_caltech101.m: simpifies the code, removes fish and vlad versions
      vl_fisher.c: bugfix
      webdoc: fixes dead links in ToCs
      dll.mak: corretly makes with DISABLE_OPENMP under Linux
      gmm: improves stability and bugfixes
      vlad.c: improves example code
      gmm.*: bugfixes
      gmm.c: fixes messages
      fisher.c: avoid NaNs on null modes
      Xcode: updates project
      generic.c: fixes two debug messages
      gmm: refactoring and bugfixes
      fisher.c: simplifies code
      vl_covdet.c: fixes bug in computation of LIOP descriptors
      adds new recognition demo
      apps: recognition: fixes auto-download of VOC data
      doc: converts help links in mfiles to work properly in the HTML documentation
      apps.html: updates table
      vl_svm.m: doc
      apps: doc and comments
      vlfeat.css: adjust sidebar
      index.html: fixes ACM link
      small fixes: compilear warnings, whitespaces, etc.
      corrects a few copyright notices
      Makefile.mak: sync
      Windows: fixes compatibility
      covdet.c: fixes crash
      doc: fixes generation of figures in Doxygen documentation

Brian Fulkerson (53):
      Makefile: isysroot syslibroot macosx-version-min for 10.5 compatibility
      sift.c: fix NaN crash when custom frames with scale 0 are passed
      sift.c: prevent log2 from going to inf when sigma = 0
      sift.c: fix typos in sift comments
      Makefile: Ignore CVPR tutorial wiki on post-doc
      vl_demo_caltech101: outline of a demo for caltech101
      vl_demo_caltech101: add tf-idf weighting, hikmeans, rename segs to ims
      make: MATLAB_EXE in generating documentation
      vl_setup: spell ready correctly
      aib: uint in appropriate places to fix warnings
      kdtree documentation updates
      mser: fix signed / unsigned warnings
      kmeans: fix signed / unsigned comparison in vl_ikmeanspush
      vl_tpsumx: fix warning, use new vlmxErrs
      hikmeans: fix signed / unsigned comparison warning
      ikmeans_elkan: fixed signed / unsigned comparison warning
      sift: Initialize keys to suppress warning
      vl_mser: initialize various pointers and dof to suppress warning
      Makefiles: builtin is BSD specific
      generic-driver: check for end of file when reading a double
      web: edits to compiling
      web: imdisttf tutorial edits
      web: correct typos on index.html
      web: edits to install-matlab instructions
      web: remove . from title in web.xml
      tutorials: add distance transform
      vl_demo: add distance transform to default set of demos
      tutorial: minor fixes to dsift tutorial
      tutorials: remove some duplicate listings of tutorials
      remove vl_demo_caltech101 in anticipation of new and improved version
      phow_caltech101: typo
      apps.html: edits
      apps: add image for caltech
      float.th: VALIGNED macro needed escaping for argument
      index: add a space and capitalize
      download.html: newer versions of git use git format-patch
      phow_caltech101: select features only from the training data
      phow_caltech101: typos in comments
      vl_imsmooth: typos in comments
      vl_binsum: typos in comments
      vl_binsum: Typo
      vl_twister: a few typos
      vl_grad: remove unused size command in order to un-break vl_grad
      using-gcc: add a simple g++ tutorial
      Makefile text cleanup
      Invert logic on VL_ENABLE_THREADS and VL_ENABLE_SSE2.
      vl-config: remove, since inverted logic takes care of most cases
      vl_siftdescriptor: add NormThresh option (thanks Michalis Raptis)
      vl_siftdescriptor: add help for new normthresh option
      vl_ubcmatch: fix case where top 2 distances are both 0 (thanks Scott Merritt)
      imdisttf, mexutils: spelling fixes from recent documentation changes
      docsrc/install-shell.html: glx => glnx32
      vl_demo_cmd.m: update for new names (glx=>glnx32, a64, glnxa64)

Daniele Perrone (26):
      First version of block sparse array
      Fix first bsarray design, First implementation of bsmatrix
      bsarray design fixed
      pegasos refactoring
      pegasos with homkermap and bsmatrix implementation
      svm refined design
      svm refactoring
      design fixes
      bug fixes and tests
      tests fix
      Documentation
      Add matlab struct 'training dataset' and adapt vl_svmpegasos to it.
      small fixes
      Matlab Documentation
      Bug fix
      Demo and tutorial on new Pegasos SVM implementation
      Fix small mistakes in the documentation
      Small Fix to pegasos demo
      clean up of old files
      more clean up
      dp bio
      fix tabs and copyright statements
      fix svmpegasos test and doc
      Fix classification demo
      Fix confusion with bias preconditioner
      liblinear fix

David (13):
      tutorials improved, added demos, documentation changes
      added demos and docs
      Added caltech demo
      Merge branch 'david2' of vlfeat.org:vlfeat into david2
      Documentation changes
      added xtests
      Documentation changes, some demo changes, \n
      added avx support for l2 distance and gmm
      vl_kdtree_query supports parallel queries, minor changes to documentation
      cosmetic changes
      demos updated
      caltech mended with new learning part, vlad parallelized
      vl_setup repaired

David Novotny (3):
      kdtree: add KDTreeSearcher to support parallel computation
      kmeans:  adds ANN and OpenMP support for ANN, Lloyd, Elkan, adds tutorial
      adds Fisher and VLAD  feature maps

Dima Kogan (13):
      removed SIFT to comply with the DFSG
      initial debianization
      added patches
      debianized GNU octave bindings
      packaged documentation
      added examples
      added README.source
      added lintian override to OK the package/library name discrepancy
      corrected debian/watch
      install the headers to /usr/include/vl/
      imported architecture-removing patch
      patch import
      changelog bump

Hana Sarbortova (4):
      adds LIOP feature descriptor
      before final one
      final
      added hana about

Karel Lenc (2):
      stringop: adds enumeration manipulation
      mexutils.h: adds vlmxDecodeEnumeration

Milan Sulc (23):
      svm: adds SDCA algorithm, move PEGASOS to SGD, better documentation
      Caltech 101 demo app: use new SVM code
      SVM: svms.h->svm.h, SVM doc in svm.h
      SVM: documentation changes
      SVM: fixing L2 conjugate function and svm.h includes, documentation
      SVM: minor changes to documentation + about
      SVM: minor changes to MATLAB documentation, Makefile.mak
      LIOP: parameters in vl_covdet.c [Hana]
      svm: changed vl_demo_svm script according to changes made to svmtrain
      svm: changed vl_demo_svm script according to changes made to svmtrain
      svm: also changed the tutorial according to changes in svmtrain/demo
      svm: vl_svmtrain typo MATLAB example fix
      svm: vl_svmtrain testing example: must set solver to NONE
      svm dataset doc: link to VL_HOMKERMAP
      svm doc typos
      vl_homkermap doc typo
      vl_svmtrain.m: fix doc (in current code, SDCA is default)
      vl_hog doc: variant name
      svm tutorial
      svm tutorial typo
      svm tutorial typo
      svm tutorial typo
      svmdataset: warning removed, vl_svmtrain: doc details

Milan Šulc (1):
      SVM: changes for compiling on Windows

Tomas Pfister (1):
      vl_roc.m: fixes typos

david (6):
      C90 compatibility issues fixed
      KD tree building speeded up by computing dimension variances on a random subset of data, minor changes to fisher caltech demo
      GMM: weights->priors, sigmas->covariances; SSE,AVX check on runtime AVX: moved to separated module, changes in host.h for AVX support, makefile changed too
      corrected private omp variables in fisher.c, gmm kmeans ownership improved, minor refractoring in matlab gmm code
      fixed c99 compatibility issues
      VLAD and FV example codes added.

-----------------------------------------------------------------------

This annotated tag includes the following new commits:

       new  5034085   patch import
       new  fa5775a   changelog bump

The 2 revisions listed above as "new" are entirely new to this
repository and will be described in separate emails.  The revisions
listed as "adds" were already present in the repository and have only
been added to this reference.


-- 
Alioth's /usr/local/bin/git-commit-notice on /srv/git.debian.org/git/debian-science/packages/vlfeat.git



More information about the debian-science-commits mailing list