This is just two algorithms, since skycoord was done previously.
Classification can't be tested without a model flux, so the unit test is bogus.

[pgee@lsst-dev mb463]$ git diff --stat origin
include/lsst/meas/base.h | 2 +
include/lsst/meas/base/Classification.h | 230 ++++++++++++++++++++
include/lsst/meas/base/PixelFlags.h | 151 +++++++++++++
include/lsst/meas/base/ResultMappers.h | 17 ++
include/lsst/meas/base/Results.h | 3 +
.../meas/base/algorithms/PixelFlagsTemplates.h | 38 ++++
python/lsst/meas/base/baseLib.i | 29 +++
python/lsst/meas/base/plugins.py | 3 +-
src/Classification.cc | 128 +++++++++++
src/PixelFlags.cc | 120 ++++++++++
tests/testClassificationBasic.py | 95 ++++++++
tests/testPixelFlagsBasic.py | 100 +++++++++
12 files changed, 915 insertions, 1 deletions

Looks good. Just a few minor comments:

Classification:

• This is a lot of boilerplate for one line of actual algorithm code; how do you feel about converting this to a pure-Python plugin?
• If you'd prefer to keep it in C++, please remove the unnecessary C standard library includes from Classification.h

PixelFlags:

• Could we just move the class in PixelFlagsTemplates.h directly into an anonymous namespace in PixelFlags.cc now? It'd be one less thing to cleanup later; I don't think PixelFlags really requires any cleanup beyond that.

Tests:

• You have a lot of variable names that don't comply with our coding standards, particularly in the use of underscores (e.g. "sfm_config"), which are not allowed, and in non-camelcase word combinations ("srccat" should be "srcCat").