Commits on Jan 2, 2009
  1. Build system tweaks to support cutting a 2.0alpha5 release.

    Matthew Hancher committed Jan 2, 2009
  2. Fixed a rather nasty bug in Rotate180PixelAccessor that would

    segfault on multi-plane images.  This was breaking the
    InterestPoint TestIntegral test_haar_filters2() test.
    Matthew Hancher committed Jan 2, 2009
  3. Be more precise when including/using memset.

    Matthew Hancher committed Jan 2, 2009
Commits on Jan 1, 2009
  1. Modified ipfind to use the latest features of VWSURF.

    Zack Moratto committed Jan 1, 2009
  2. Adding implementations of FH15 and SURF128 again. Added more tests for

    SURF, specifically Haar filters, subpixel refinement, orientation 
    assignment. Haar Filters now use template ImageViewBase if the user wants to 
    provide an InterpolatedView of the integral image.
    Zack Moratto committed Jan 1, 2009
  3. Mistake in the descriptor

    Zack Moratto committed Jan 1, 2009
  4. Incorporate more matrix/quaternion conversion tests.

    Matthew Hancher committed Jan 1, 2009
  5. Removed the deprecated apply_per_pixel function. Use for_each_pixel i…

    Matthew Hancher committed Jan 1, 2009
  6. Convert the in-place pixel math to use for_each_pixel instead of appl…

    Matthew Hancher committed Jan 1, 2009
  7. Extend for_each_pixel() to support ternary functions.

    Matthew Hancher committed Jan 1, 2009
  8. Added multi_solve_symmetric for solution of equations A*X = B where X…

    … and B are matrices, not simply vectors. Added test case to.
    Liam Pedersen committed Jan 1, 2009
Commits on Dec 29, 2008
  1. Allow Ransac functor and template class to be templated and work with…

    … point
    sets of different type.
    Ransac can also fit samples of different types (i.e. 3-D points to 2-D points)
    as long as the Fitting and Error functors convert them accordingly and the
    templating allows for different point types; which it does now.
    Simon Rutishauser committed Dec 29, 2008
Commits on Dec 25, 2008
Commits on Dec 24, 2008
  1. Added checking of descriptor generator and interest operator strings

    before starting the job. (This is really only a problem with the 
    descriptor string, the code would have finished finding interest points 
    then crashed when trying to calculate descriptor) I also believe this is 
    what pedersen was referencing in Ticker #5.
    Zack Moratto committed Dec 24, 2008
Commits on Dec 23, 2008
  1. Changed to work with the new SURF style (Integral Images are handed

    externally in Ipfind now ).
    Zack Moratto committed Dec 23, 2008
  2. Changing up SURF descriptor quite a bit. It now uses the integral ima…

    …ge and is
    threaded. Yet, unfortunately no longer fits into the 
    DescriptorGeneratorBase model. Also wrote some tests to integral image.
    Zack Moratto committed Dec 23, 2008
Commits on Dec 22, 2008
  1. make the qtree default callbacks be functors, to work around the gcc 4.3

    / default arguments problem. This resolves #134.
    novas0x2a committed Dec 22, 2008
Commits on Dec 20, 2008
  1. polished the code coverage some- filter out paths that aren't ours, a…

    …nd do
    an initial pass on the source to make sure it finds all code. 57%
    coverage. Ouch.
    novas0x2a committed Dec 20, 2008
  2. remove blank line

    novas0x2a committed Dec 20, 2008
  3. add support for generating a test coverage file. pass

    --enable-test-coverage to configure, then run make check and then make
    lcov. You'll need to install lcov. It'll generate an index.html in
    novas0x2a committed Dec 20, 2008
  4. Copying an uninitialized (singular) iterator is undefined. Here's a

    workaround for that.
    novas0x2a committed Dec 20, 2008
Commits on Dec 19, 2008
  1. give the relicense script some love. it's all ready to go when we figure

    out the copyright stuff.
    novas0x2a committed Dec 19, 2008
  2. move some stuff around

    novas0x2a committed Dec 19, 2008
  3. remooooooove the graaaaaveyard

    novas0x2a committed Dec 19, 2008
  4. Added LAPACK QR and RQ decomposition wrapper methods including some t…

    Simon Rutishauser committed Dec 19, 2008
Commits on Dec 18, 2008
  1. Disabled SURF128, acting funny. Cleaned up some of the other code. Ne…

    …ed to start using integral image for
    Zack Moratto committed Dec 18, 2008
  2. this adds celestia "virtual texture" support to image2qtree. It doesn…

    …'t do
    terrain / normal maps, though celestia supports it. Next version, maybe.
    novas0x2a committed Dec 18, 2008
Commits on Dec 17, 2008
  1. Added threading to the SURF detector. The detector has also been renamed

    FH9, as eventually I'll finish writing FH15. Still working on threading 
    the descriptor.
    Zack Moratto committed Dec 17, 2008
  2. make it so AX_PKG supports requiring specific functions, and use that to

    require a libGL with glsl support. the mesa on obelix is full of lies, and
    claims opengl 2.0 support in the header without support in the lib.
    novas0x2a committed Dec 17, 2008
Commits on Dec 16, 2008
  1. warning fixes: variable possibly undefined, use size_t instead of lon…

    …g for
    return of ftell, check fread return code, add missing {}, remove unused
    variable, make sure g++ knows that the open check isn't dead code, and
    remove impossible range checks
    novas0x2a committed Dec 16, 2008
  2. AS_VERSION_COMPARE is new in 2.60. Backport it (inside ifdefs) and re…

    …move autotroll's hard-coded 2.60 dep to unbreak the build on some of the irg machines
    novas0x2a committed Dec 16, 2008
Commits on Dec 15, 2008
  1. Major overhaul of SURF algorithm. I hadn't noticed a paper came out in

    September 2008 about SURF. It was very enlighting. Results are now much 
    closer to SURF's results but some fine tuning is still needed. Note: VW 
    currently only has SURF's FH9 detector. The more accurate scale space 
    version FH15 I'll have out after christmas break. Also, I've added 
    SURF128 descriptor.
    Zack Moratto committed Dec 15, 2008
Commits on Dec 12, 2008
  1. Attempt at a fix for Vinh\'s OpenGL linker problem. Not sure why this…

    … doesn\'t happen on my machine, but this seems to be the Apple-approved method of fixing it.
    broxtronix committed Dec 12, 2008