def testClassificationFrame(self):
handDetectorFile = DATASET_HANDDETECTOR_GT_PATH.format(self.dataset,
self.feature,
self.classifier,
self.dynamic)
if self.dynamic == "":
self.handDetector = ObjectPickler.load(HandDetector, handDetectorFile)
else:
self.handDetector = ObjectPickler.load(DynamicHandDetector, handDetectorFile)
self.handDetector.setOptimalParameters(50.0)
frame = ObjectPickler.load(Frame, VIDEO_EXAMPLE_PATH.format("frameMatrix.pk"))
hands = self.handDetector.classifyFrame(frame,dtype="integer")
self.assertIsInstance(hands, numpy.ndarray,
msg="Hand detector is not returning arrays")
self.assertEqual(hands.size, 1,
msg="Hand detector is not returning arrays of size 1")
self.assertIsInstance(hands[0], int,
msg="Hand detector is not returning arrays of size 1 with an integer")
hands = self.handDetector.classifyFrame(frame,dtype="float")
self.assertIsInstance(hands, numpy.ndarray,
msg="Hand detector is not returning arrays")
self.assertEqual(hands.size, 1,
msg="Hand detector is not returning arrays of size 1")
self.assertIsInstance(hands[0], float,
msg="Hand detector is not returning arrays of size 1 with a float")
def runTest(self):
frame = ObjectPickler.load(Frame,
VIDEO_EXAMPLE_PATH.format("frameMatrix.pk"))
success, desc = self.featureController.getFeatures(frame)
desc = desc.next()
if createGroundTruths:
gtfile = GROUNDTRUTH_FEATURE_PATH.format(self.feature)
print "[Feature Creator] Ground Truth Created"
print gtfile
if not os.path.exists(os.path.split(gtfile)[0]):
os.makedirs(os.path.split(gtfile)[0])
ObjectPickler.save(desc, gtfile)
self.assertIsInstance(desc, numpy.ndarray)
desc2 = ObjectPickler.load(
Feature, GROUNDTRUTH_FEATURE_PATH.format(self.feature))
numpy.testing.assert_equal(desc, desc2)
def frameExportImport(self):
videoname = VIDEO_EXAMPLE_PATH.format(self.videoname)
video = Video(videoname)
success, frame = video.read()
video.release()
filename = VIDEO_EXAMPLE_PATH.format("frameMatrix.pk")
ObjectPickler.save(frame, filename)
frame2 = ObjectPickler.load(Frame, filename)
self.assertEqual(frame,
frame2,
msg="Readed frame different of ground truth")
def testClassificationVideo(self):
handDetectorFile = DATASET_HANDDETECTOR_GT_PATH.format(self.dataset,
self.feature,
self.classifier,
self.dynamic)
self.handDetector = ObjectPickler.load(HandDetector, handDetectorFile)
video = Video(VIDEO_EXAMPLE_PATH.format("BENCHTEST.MP4"))
hands = self.handDetector.classifyVideo(video,dtype="integer")
self.assertIsInstance(hands, numpy.ndarray,
msg="Hand detector is not returning arrays")
video.release()
def testTrainingDetector(self):
datasetFolder = DATASET_DATAMANAGER_GT_PATH.format(self.dataset,self.feature)
dm = ObjectPickler.load(HandDetectionDataManager, datasetFolder)
self.handDetector.trainClassifier(dm)
handDetectorFile = DATASET_HANDDETECTOR_GT_PATH.format(self.dataset,
self.feature,
self.classifier,
self.dynamic)
if createGroundTruths:
print "[Hand Detector] Ground Truth Created"
print handDetectorFile
if not os.path.exists(os.path.split(handDetectorFile)[0]):
os.makedirs(os.path.split(handDetectorFile)[0])
success = ObjectPickler.save(self.handDetector, gtfile)
if self.dynamic == "":
hd = ObjectPickler.load(HandDetector, handDetectorFile)
else:
hd = ObjectPickler.load(DynamicHandDetector, handDetectorFile)
hd.setOptimalParameters(50.0)
featureVideoPath = GROUNDTRUTH_VIDEOFEATURE_PATH.format(self.videotest, self.feature)
featureVideo = ObjectPickler.load(FeatureVideo, featureVideoPath)
featureVideo.features = featureVideo.features[:500]
r1 = self.handDetector.classifyFeatureVideo(featureVideo,dtype="integer")
r2 = hd.classifyFeatureVideo(featureVideo,dtype="integer")
numpy.testing.assert_array_almost_equal(r1,r2,)
if self.classifier == "SVM":
if self.dynamic == "_dynamic": #restart the dynamic properties
self.handDetector = DynamicHandDetector(self.feature, 180,"SVM")
self.handDetector.trainClassifier(dm)
r3 = self.handDetector.classifyFeatureVideo(featureVideo,dtype='float')
r4 = self.handDetector.binarizeDetections(r3,th=0)
numpy.testing.assert_array_almost_equal(r1,r4)
self.assertIsInstance(r1, numpy.ndarray,
msg="Trained Hand detector is not returning arrays")
self.assertIsInstance(r2, numpy.ndarray,
msg="Ground truth Hand detector is not returning arrays")
def testDataManager(self):
datasetFolder = DATASET_PATH.format(self.dataset)
if os.path.exists(datasetFolder):
dm = HandDetectionDataManager()
dm.readDataset(datasetFolder,180,feature=self.feature)
if createGroundTruths:
print "[Hand Detection Data Manager] Ground Truth Created"
gtfile = DATASET_DATAMANAGER_GT_PATH.format(self.dataset,self.feature)
print gtfile
if not os.path.exists(os.path.split(gtfile)[0]):
os.makedirs(os.path.split(gtfile)[0])
success = ObjectPickler.save(dm, gtfile)
dm2 = ObjectPickler.load(HandDetectionDataManager, DATASET_DATAMANAGER_GT_PATH.format(self.dataset,self.feature))
numpy.testing.assert_equal(dm.attributes,dm2.attributes)
self.assertIsInstance(dm.attributes, numpy.ndarray,
msg="New HandDetectionDataManager does not guarantee ndarray")
self.assertIsInstance(dm2.attributes, numpy.ndarray,
msg="GroundTruth is not ndarray")
else:
raise os.error("The dataset does not exists")
def runTest(self):
mask = DATASET_MASKS_PATH.format(dataset, trainingVideo) + trainingMask
gtfile = DATASET_HSDATAMANAGER_GT_PATH.format(dataset, trainingVideo,
trainingMask,
self.feature)
dm = MaskBasedDataManager()
dm.readDataset([mask], 200, "LAB")
if createGroundTruths:
print "[Feature Creator] Ground Truth Created"
print gtfile
if not os.path.exists(os.path.split(gtfile)[0]):
os.makedirs(os.path.split(gtfile)[0])
ObjectPickler.save(dm, gtfile)
dm2 = ObjectPickler.load(MaskBasedDataManager, gtfile)
numpy.testing.assert_equal(dm.attributes, dm2.attributes)
self.assertIsInstance(
dm.attributes,
numpy.ndarray,
msg="New HandDetectionDataManager does not guarantee ndarray")
self.assertIsInstance(dm2.attributes,
numpy.ndarray,
msg="GroundTruth is not ndarray")
def testVideoFeatureLoader(self):
pickleName = GROUNDTRUTH_VIDEOFEATURE_PATH.format(
self.videoname, self.feature)
featureVideo = ObjectPickler.load(FeatureVideo, pickleName)
self.assertIsInstance(featureVideo.features, numpy.ndarray)