Mercurial Hosting > traffic-intelligence

comparison python/moving.py @ 898:1fc901d983ed

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
better take into account unknown appearance classification
author Nicolas Saunier <nicolas.saunier@polymtl.ca>
date Wed, 21 Jun 2017 17:21:03 -0400
parents ff92801e5c54
children 8f60ecfc2f06
comparison
equal deleted inserted replaced
897:f5a49b603e8b 898:1fc901d983ed
1588 hog = cvutils.HOG(croppedImg, rescaleSize, orientations, pixelsPerCell, cellsPerBlock, blockNorm, visualize=False, normalize=False) 1588 hog = cvutils.HOG(croppedImg, rescaleSize, orientations, pixelsPerCell, cellsPerBlock, blockNorm, visualize=False, normalize=False)
1589 self.userTypes[instant] = int(self.appearanceClassifier.predict(hog)) 1589 self.userTypes[instant] = int(self.appearanceClassifier.predict(hog))
1590 else: 1590 else:
1591 self.userTypes[instant] = userType2Num['unknown'] 1591 self.userTypes[instant] = userType2Num['unknown']
1592 1592
1593 def classifyUserTypeHoGSVM(self, pedBikeCarSVM = None, width = 0, height = 0, homography = None, images = None, bikeCarSVM = None, pedBikeSpeedTreshold = float('Inf'), bikeCarSpeedThreshold = float('Inf'), minSpeedEquiprobable = -1, speedProbabilities = None, aggregationFunc = median, nInstantsIgnoredAtEnds = 0, px = 0.2, py = 0.2, minNPixels = 800, rescaleSize = (64, 64), orientations = 9, pixelsPerCell = (8,8), cellsPerBlock = (2,2)): 1593 def classifyUserTypeHoGSVM(self, pedBikeCarSVM = None, width = 0, height = 0, homography = None, images = None, bikeCarSVM = None, pedBikeSpeedTreshold = float('Inf'), bikeCarSpeedThreshold = float('Inf'), minSpeedEquiprobable = -1, speedProbabilities = None, aggregationFunc = median, maxPercentUnknown = 0.5, nInstantsIgnoredAtEnds = 0, px = 0.2, py = 0.2, minNPixels = 800, rescaleSize = (64, 64), orientations = 9, pixelsPerCell = (8,8), cellsPerBlock = (2,2)):
1594 '''Agregates SVM detections in each image and returns probability 1594 '''Agregates SVM detections in each image and returns probability
1595 (proportion of instants with classification in each category) 1595 (proportion of instants with classification in each category)
1596 1596
1597 images is a dictionary of images indexed by instant 1597 images is a dictionary of images indexed by instant
1598 With default parameters, the general (ped-bike-car) classifier will be used 1598 With default parameters, the general (ped-bike-car) classifier will be used
1609 # compute P(Speed|Class) 1609 # compute P(Speed|Class)
1610 if speedProbabilities is None or self.aggregatedSpeed < minSpeedEquiprobable: # equiprobable information from speed 1610 if speedProbabilities is None or self.aggregatedSpeed < minSpeedEquiprobable: # equiprobable information from speed
1611 userTypeProbabilities = {userType2Num['car']: 1., userType2Num['pedestrian']: 1., userType2Num['bicycle']: 1.} 1611 userTypeProbabilities = {userType2Num['car']: 1., userType2Num['pedestrian']: 1., userType2Num['bicycle']: 1.}
1612 else: 1612 else:
1613 userTypeProbabilities = {userType2Num[userTypename]: speedProbabilities[userTypename](self.aggregatedSpeed) for userTypename in speedProbabilities} 1613 userTypeProbabilities = {userType2Num[userTypename]: speedProbabilities[userTypename](self.aggregatedSpeed) for userTypename in speedProbabilities}
1614 # compute P(Class|Appearance)
1615 nInstantsUserType = {userTypeNum: 0 for userTypeNum in userTypeProbabilities}# number of instants the object is classified as userTypename
1616 nInstantsUserType[userType2Num['unknown']] = 0
1617 for t in self.userTypes:
1618 nInstantsUserType[self.userTypes[t]] += 1 #nInstantsUserType.get(self.userTypes[t], 0) + 1
1614 # result is P(Class|Appearance) x P(Speed|Class) 1619 # result is P(Class|Appearance) x P(Speed|Class)
1615 nInstantsUserType = {userTypeNum: 0 for userTypeNum in userTypeProbabilities}# number of instants the object is classified as userTypename 1620 if nInstantsUserType[userType2Num['unknown']] < maxPercentUnknown*self.length(): # if not too many unknowns
1616 for t in self.userTypes: 1621 for userTypeNum in userTypeProbabilities:
1617 nInstantsUserType[self.userTypes[t]] = nInstantsUserType.get(self.userTypes[t], 0) + 1 1622 userTypeProbabilities[userTypeNum] *= nInstantsUserType[userTypeNum]
1618 for userTypeNum in userTypeProbabilities:
1619 userTypeProbabilities[userTypeNum] *= nInstantsUserType[userTypeNum]
1620 # class is the user type that maximizes usertype probabilities 1623 # class is the user type that maximizes usertype probabilities
1621 self.setUserType(utils.argmaxDict(userTypeProbabilities)) 1624 if nInstantsUserType[userType2Num['unknown']] >= maxPercentUnknown*self.length() and (speedProbabilities is None or self.aggregatedSpeed < minSpeedEquiprobable): # if no speed information and too many unknowns
1625 self.setUserType(userType2Num['unknown'])
1626 else:
1627 self.setUserType(utils.argmaxDict(userTypeProbabilities))
1622 1628
1623 def classifyUserTypeArea(self, areas, homography): 1629 def classifyUserTypeArea(self, areas, homography):
1624 '''Classifies the object based on its location (projected to image space) 1630 '''Classifies the object based on its location (projected to image space)
1625 areas is a dictionary of matrix of the size of the image space 1631 areas is a dictionary of matrix of the size of the image space
1626 for different road users possible locations, indexed by road user type names 1632 for different road users possible locations, indexed by road user type names