else:
c = ev.Contender("DPM")
c.detectorString = "DPM_Detector"
c.detectorData = "detectors/dpmStarbucksLogo.zip"
c.foundMap = {'Positive':easy.getPurpose('pos'), 'Negative':easy.getPurpose('neg')}
contenders.append( c );
# OpenCVCascade, with special settings for anticipated poor performance
if (easy.getTrainer("OpenCVCascadeTrainer")==None):
print("Cascade service(s) are insufficiently configured, skipping.")
else:
c = ev.Contender("cascade")
c.trainerString = "OpenCVCascadeTrainer"
c.detectorString = "OpenCVCascadeDetector"
# c.foundMap = {'any':easy.getPurpose('pos')}
c.foundMap = {'positive':posPurpose, 'negative':negPurpose}
detector = easy.getDetector(c.detectorString)
detectorProps = easy.getDetectorProperties(detector)
c.detectorProps = detectorProps;
c.detectorProps.props["maxRectangles"] = "200"
c.detectorProps.minNeighbors = 0; # This prevents hang up in evaluator when training has too few samples
contenders.append( c );
runset = easy.createRunSet( "trainImg/kr", "pos" )
easy.addToRunSet( runset, "trainImg/ca", "neg" )
easy.printRunSetInfo( runset, printArtifacts=False, printLabels=True )
perfdata = ev.joust( contenders, runset, folds=3 )
ev.printEvaluationResults(perfdata[0])
print("BOW detector service is insufficiently configured, skipping.")
else:
c = ev.Contender("BOW")
c.detectorString = "BOW_Detector"
c.detectorData = "detectors/bowUSKOCA.zip"
c.foundMap = {
'kr': easy.getPurpose('pos'),
'ca': easy.getPurpose('neg'),
'us': easy.getPurpose('neg'),
'unlabeled': easy.getPurpose('neg')
}
contenders.append(c)
# OpenCV Cascade detector
if (easy.getDetector("OpenCVCascadeDetector") == None):
print(
"OpenCVCascadeDetector service is insufficiently configured, skipping."
)
else:
c = ev.Contender("Faces")
c.detectorString = "OpenCVCascadeDetector"
c.detectorData = "detectors/OpencvFaces.zip"
c.foundMap = {
'positive': easy.getPurpose('pos'),
'negative': easy.getPurpose('neg')
}
contenders.append(c)
perfdata = ev.joust(contenders, runset)
ev.printEvaluationResults(perfdata[0])
c.detectorData = "detectors/dpmStarbucksLogo.zip"
c.foundMap = {
'Positive': easy.getPurpose('pos'),
'Negative': easy.getPurpose('neg')
}
contenders.append(c)
# OpenCVCascade, with special settings for anticipated poor performance
if (easy.getTrainer("OpenCVCascadeTrainer") == None):
print("Cascade service(s) are insufficiently configured, skipping.")
else:
c = ev.Contender("cascade")
c.trainerString = "OpenCVCascadeTrainer"
c.detectorString = "OpenCVCascadeDetector"
# c.foundMap = {'any':easy.getPurpose('pos')}
c.foundMap = {'positive': posPurpose, 'negative': negPurpose}
detector = easy.getDetector(c.detectorString)
detectorProps = easy.getDetectorProperties(detector)
c.detectorProps = detectorProps
c.detectorProps.props["maxRectangles"] = "200"
c.detectorProps.minNeighbors = 0
# This prevents hang up in evaluator when training has too few samples
contenders.append(c)
runset = easy.createRunSet("trainImg/kr", "pos")
easy.addToRunSet(runset, "trainImg/ca", "neg")
easy.printRunSetInfo(runset, printArtifacts=False, printLabels=True)
perfdata = ev.joust(contenders, runset, folds=3)
ev.printEvaluationResults(perfdata[0])
#easy.printRunSetInfo( runset, printArtifacts=False, printLabels=True )
contenders = []
# Bag of Words
if (easy.getDetector("BOW_Detector")==None):
print("BOW detector service is insufficiently configured, skipping.")
else:
c = ev.Contender("BOW")
c.detectorString = "BOW_Detector"
c.detectorData = "detectors/bowUSKOCA.zip"
c.foundMap = { 'kr':easy.getPurpose('pos'),
'ca':easy.getPurpose('neg'),
'us':easy.getPurpose('neg'),
'unlabeled':easy.getPurpose('neg')}
contenders.append( c );
# OpenCV Cascade detector
if (easy.getDetector("OpenCVCascadeDetector")==None):
print("OpenCVCascadeDetector service is insufficiently configured, skipping.")
else:
c = ev.Contender("Faces")
c.detectorString = "OpenCVCascadeDetector"
c.detectorData = "detectors/OpencvFaces.zip"
c.foundMap = { 'positive':easy.getPurpose('pos'),
'negative':easy.getPurpose('neg')}
contenders.append( c );
perfdata = ev.joust( contenders, runset )
ev.printEvaluationResults(perfdata[0])
