Skip to content

Commit

Permalink
Adapt to change in meas_base test dataset generation code.
Browse files Browse the repository at this point in the history
  • Loading branch information
@TallJimbo
TallJimbo committed Mar 17, 2015
1 parent 8be98ec commit 584bbdc
Showing 1 changed file with 49 additions and 63 deletions.
112 changes: 49 additions & 63 deletions tests/testCModelPlugins.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -27,12 +27,10 @@
import lsst.afw.geom
import lsst.afw.table
import lsst.utils.tests
import lsst.meas.base.tests
import lsst.meas.modelfit

import lsst.afw.display
from lsst.meas.base.tests import AlgorithmTestCase, TestDataset

numpy.random.seed(1234567)

# n.b. Some tests here depend on the noise realization in the test data
# or from the numpy random number generator.
Expand All @@ -42,81 +40,69 @@
# the measured flux lies within 2 sigma of the correct value, which we
# should expect to fail sometimes.

class CModelTestCase(lsst.meas.base.tests.AlgorithmTestCase):
class CModelTestCase(AlgorithmTestCase):
"""Test case for the CModel measurement plugins
"""

def setUp(self):
lsst.meas.base.tests.AlgorithmTestCase.setUp(self)
self.record = self.truth[0]
AlgorithmTestCase.setUp(self)
self.bbox = lsst.afw.geom.Box2I(lsst.afw.geom.Point2I(0, 0),
lsst.afw.geom.Extent2I(200, 100))
self.dataset = TestDataset(self.bbox)
# first source is a point
self.dataset.addSource(100000.0, lsst.afw.geom.Point2D(50.1, 49.8))
# second source is extended
self.dataset.addSource(100000.0, lsst.afw.geom.Point2D(149.9, 50.3),
lsst.afw.geom.ellipses.Quadrupole(8, 9, 3))

def tearDown(self):
lsst.meas.base.tests.AlgorithmTestCase.tearDown(self)
del self.record
AlgorithmTestCase.tearDown(self)
del self.bbox
del self.dataset

def checkOutputs(self, measCat):
def checkOutputs(self, measCat, truthCat=None):
"""Test that the outputs of the CModel plugins are reasonable, and that the bookkeeping works.
Science-quality tests either in testCModel.py (where we call the same code via a different interface)
or something we have to do statistically on real data.
"""
for measRecord, truthRecord in zip(measCat, self.truth):
if truthCat is None: truthCat = measCat
for measRecord, truthRecord in zip(measCat, truthCat):
trueFlux = truthRecord.get("truth_flux")
if not measRecord.getShapeFlag():
self.assertFalse(measRecord.get("modelfit_CModel_initial_flag"))
self.assertFalse(measRecord.get("modelfit_CModel_exp_flag"))
self.assertFalse(measRecord.get("modelfit_CModel_dev_flag"))
self.assertFalse(measRecord.get("modelfit_CModel_flag"))
self.assertClose(measRecord.get("modelfit_CModel_flux"), trueFlux, rtol=0.5)
self.assertGreater(measRecord.get("modelfit_CModel_fluxSigma"), 0.0)
self.assertClose(measRecord.get("modelfit_CModel_initial_flux"), trueFlux, rtol=0.5)
self.assertGreater(measRecord.get("modelfit_CModel_initial_fluxSigma"), 0.0)
self.assertClose(measRecord.get("modelfit_CModel_exp_flux"), trueFlux, rtol=0.5)
self.assertGreater(measRecord.get("modelfit_CModel_exp_fluxSigma"), 0.0)
self.assertClose(measRecord.get("modelfit_CModel_dev_flux"), trueFlux, rtol=0.5)
self.assertGreater(measRecord.get("modelfit_CModel_dev_fluxSigma"), 0.0)
else:
self.assertTrue(measRecord.get("modelfit_CModel_initial_flag"))
self.assertTrue(measRecord.get("modelfit_CModel_exp_flag"))
self.assertTrue(measRecord.get("modelfit_CModel_dev_flag"))
self.assertTrue(measRecord.get("modelfit_CModel_flag"))
self.assertFalse(measRecord.get("modelfit_CModel_initial_flag"))
self.assertFalse(measRecord.get("modelfit_CModel_exp_flag"))
self.assertFalse(measRecord.get("modelfit_CModel_dev_flag"))
self.assertFalse(measRecord.get("modelfit_CModel_flag"))
self.assertClose(measRecord.get("modelfit_CModel_flux"), trueFlux, rtol=0.5)
self.assertGreater(measRecord.get("modelfit_CModel_fluxSigma"), 0.0)
self.assertClose(measRecord.get("modelfit_CModel_initial_flux"), trueFlux, rtol=0.5)
self.assertGreater(measRecord.get("modelfit_CModel_initial_fluxSigma"), 0.0)
self.assertClose(measRecord.get("modelfit_CModel_exp_flux"), trueFlux, rtol=0.5)
self.assertGreater(measRecord.get("modelfit_CModel_exp_fluxSigma"), 0.0)
self.assertClose(measRecord.get("modelfit_CModel_dev_flux"), trueFlux, rtol=0.5)
self.assertGreater(measRecord.get("modelfit_CModel_dev_fluxSigma"), 0.0)

def testPlugins(self):
# Start with a run on some simulated data, using the single-frame measurement driver
sfmConfig = lsst.meas.base.SingleFrameMeasurementTask.ConfigClass()
sfmConfig.plugins.names = ["base_SdssShape", "base_PsfFlux", "modelfit_CModel",
"modelfit_ShapeletPsfApprox"]
sfmConfig.slots.centroid = None
sfmConfig.slots.shape = "base_SdssShape"
sfmConfig.slots.psfFlux = "base_PsfFlux"
sfmConfig.slots.apFlux = None
sfmConfig.slots.modelFlux = None
sfmConfig.slots.instFlux = None
sfmSchemaMapper = lsst.afw.table.SchemaMapper(self.truth.schema)
sfmSchemaMapper.addMinimalSchema(self.truth.schema)
sfmTask = lsst.meas.base.SingleFrameMeasurementTask(config=sfmConfig,
schema=sfmSchemaMapper.editOutputSchema())
sfmMeasCat = lsst.afw.table.SourceCatalog(sfmSchemaMapper.getOutputSchema())
sfmMeasCat.extend(self.truth, sfmSchemaMapper)
sfmMeasCat.table.defineCentroid("truth")
sfmTask.run(sfmMeasCat, self.calexp)
self.checkOutputs(sfmMeasCat)

# Now we use the SFM results as the reference catalog for a forced measurement run
forcedConfig = lsst.meas.base.ForcedMeasurementTask.ConfigClass()
forcedConfig.plugins.names = ["base_TransformedCentroid", "base_TransformedShape",
"base_PsfFlux", "modelfit_CModel", "modelfit_ShapeletPsfApprox"]
forcedConfig.slots.centroid = "base_TransformedCentroid"
forcedConfig.slots.shape = "base_TransformedShape"
forcedConfig.slots.psfFlux = "base_PsfFlux"
forcedConfig.slots.apFlux = None
forcedConfig.slots.modelFlux = None
forcedConfig.slots.instFlux = None
forcedTask = lsst.meas.base.ForcedMeasurementTask(config=forcedConfig, refSchema=sfmMeasCat.schema)
forcedMeasCat = forcedTask.run(self.calexp, sfmMeasCat, self.calexp.getWcs()).sources
# only test the first three simulated sources: the last two are deblend children, and because
# of the lack of heavy footprints in forced mode, we don't measure those well at all.
self.checkOutputs(forcedMeasCat[:3])
"""Test that the plugin for single-frame measurement works, then use those outputs to test
that the forced measurement plugin works."""
plugin = "modelfit_CModel"
dependencies = ("modelfit_ShapeletPsfApprox", "base_PsfFlux")
sfmTask = self.makeSingleFrameMeasurementTask(plugin, dependencies=dependencies)
forcedTask = self.makeForcedMeasurementTask(plugin, dependencies=dependencies,
refSchema=sfmTask.schema)
# catalog1 will contain both the SFM outputs and the truth catalog for sources in exposure 1.
# Those SFM outputs will also be used as the references for the forced task.
exposure1, catalog1 = self.dataset.realize(10.0, sfmTask.schema)
sfmTask.run(catalog1, exposure1)
self.checkOutputs(catalog1)
wcs2 = self.dataset.makePerturbedWcs(self.dataset.exposure.getWcs())
dataset2 = self.dataset.transform(wcs2)
# catalog2 will contain only the truth catalog for sources in exposure 1; the structure of
# ForcedMeasurementTask means we can't put the outputs in the same catalog.
exposure2, catalog2 = dataset2.realize(10.0, dataset2.makeMinimalSchema())
results = forcedTask.run(exposure2, refCat=catalog1, refWcs=exposure1.getWcs())
self.checkOutputs(results.sources, catalog2)


def suite():
"""Returns a suite containing all the test cases in this module."""
Expand Down

0 comments on commit 584bbdc

Please sign in to comment.