Mercurial Hosting > traffic-intelligence
changeset 993:e8eabef7857c
update to OpenCV3 for python
| author | Nicolas Saunier <nicolas.saunier@polymtl.ca> |
|---|---|
| date | Wed, 16 May 2018 21:06:52 -0400 |
| parents | 2cd1ce245024 |
| children | 8118c6b77d7c |
| files | python/cvutils.py python/ml.py python/moving.py scripts/classify-objects.py scripts/train-object-classification.py |
| diffstat | 5 files changed, 85 insertions(+), 95 deletions(-) [+] |
line wrap: on
line diff
--- a/python/cvutils.py Fri Apr 13 16:48:02 2018 -0400 +++ b/python/cvutils.py Wed May 16 21:06:52 2018 -0400 @@ -94,15 +94,6 @@ H, mask = cv2.findHomography(srcPoints, dstPoints, method, ransacReprojThreshold) return H - def arrayToCvMat(a, t = cv2.CV_64FC1): - '''Converts a numpy array to an OpenCV CvMat, with default type CV_64FC1.''' - print('Deprecated, use new interface') - cvmat = cv2.cv.CreateMat(a.shape[0], a.shape[1], t) - for i in range(cvmat.rows): - for j in range(cvmat.cols): - cvmat[i,j] = a[i,j] - return cvmat - def cvPlot(img, positions, color, lastCoordinate = None, **kwargs): if lastCoordinate is None: last = positions.length()-1 @@ -152,7 +143,7 @@ nFramesShown = 0 if firstFrameNums is not None: for i in xrange(len(captures)): - captures[i].set(cv2.cv.CV_CAP_PROP_POS_FRAMES, firstFrameNums[i]) + captures[i].set(cv2.PROP_POS_FRAMES, firstFrameNums[i]) while ret and not quitKey(key): rets = [] images = [] @@ -174,45 +165,45 @@ nFramesShown += step if step > 1: for i in xrange(len(captures)): - captures[i].set(cv2.cv.CV_CAP_PROP_POS_FRAMES, firstFrameNums[i]+nFramesShown) + captures[i].set(cv2.CAP_PROP_POS_FRAMES, firstFrameNums[i]+nFramesShown) cv2.destroyAllWindows() else: print('Video captures for {} failed'.format(filenames)) def infoVideo(filename): '''Provides all available info on video ''' - cvPropertyNames = {cv2.cv.CV_CAP_PROP_FORMAT: "format", - cv2.cv.CV_CAP_PROP_FOURCC: "codec (fourcc)", - cv2.cv.CV_CAP_PROP_FPS: "fps", - cv2.cv.CV_CAP_PROP_FRAME_COUNT: "number of frames", - cv2.cv.CV_CAP_PROP_FRAME_HEIGHT: "heigh", - cv2.cv.CV_CAP_PROP_FRAME_WIDTH: "width", - cv2.cv.CV_CAP_PROP_RECTIFICATION: "rectification", - cv2.cv.CV_CAP_PROP_SATURATION: "saturation"} + cvPropertyNames = {cv2.CAP_PROP_FORMAT: "format", + cv2.CAP_PROP_FOURCC: "codec (fourcc)", + cv2.CAP_PROP_FPS: "fps", + cv2.CAP_PROP_FRAME_COUNT: "number of frames", + cv2.CAP_PROP_FRAME_HEIGHT: "heigh", + cv2.CAP_PROP_FRAME_WIDTH: "width", + cv2.CAP_PROP_RECTIFICATION: "rectification", + cv2.CAP_PROP_SATURATION: "saturation"} capture = cv2.VideoCapture(filename) videoProperties = {} if capture.isOpened(): - for cvprop in [#cv2.cv.CV_CAP_PROP_BRIGHTNESS - #cv2.cv.CV_CAP_PROP_CONTRAST - #cv2.cv.CV_CAP_PROP_CONVERT_RGB - #cv2.cv.CV_CAP_PROP_EXPOSURE - cv2.cv.CV_CAP_PROP_FORMAT, - cv2.cv.CV_CAP_PROP_FOURCC, - cv2.cv.CV_CAP_PROP_FPS, - cv2.cv.CV_CAP_PROP_FRAME_COUNT, - cv2.cv.CV_CAP_PROP_FRAME_HEIGHT, - cv2.cv.CV_CAP_PROP_FRAME_WIDTH, - #cv2.cv.CV_CAP_PROP_GAIN, - #cv2.cv.CV_CAP_PROP_HUE - #cv2.cv.CV_CAP_PROP_MODE - #cv2.cv.CV_CAP_PROP_POS_AVI_RATIO - #cv2.cv.CV_CAP_PROP_POS_FRAMES - #cv2.cv.CV_CAP_PROP_POS_MSEC - #cv2.cv.CV_CAP_PROP_RECTIFICATION, - #cv2.cv.CV_CAP_PROP_SATURATION + for cvprop in [#cv2.CAP_PROP_BRIGHTNESS + #cv2.CAP_PROP_CONTRAST + #cv2.CAP_PROP_CONVERT_RGB + #cv2.CAP_PROP_EXPOSURE + cv2.CAP_PROP_FORMAT, + cv2.CAP_PROP_FOURCC, + cv2.CAP_PROP_FPS, + cv2.CAP_PROP_FRAME_COUNT, + cv2.CAP_PROP_FRAME_HEIGHT, + cv2.CAP_PROP_FRAME_WIDTH, + #cv2.CAP_PROP_GAIN, + #cv2.CAP_PROP_HUE + #cv2.CAP_PROP_MODE + #cv2.CAP_PROP_POS_AVI_RATIO + #cv2.CAP_PROP_POS_FRAMES + #cv2.CAP_PROP_POS_MSEC + #cv2.CAP_PROP_RECTIFICATION, + #cv2.CAP_PROP_SATURATION ]: prop = capture.get(cvprop) - if cvprop == cv2.cv.CV_CAP_PROP_FOURCC and prop > 0: + if cvprop == cv2.CAP_PROP_FOURCC and prop > 0: prop = int2FOURCC(int(prop)) videoProperties[cvPropertyNames[cvprop]] = prop else: @@ -224,12 +215,12 @@ images = [] capture = cv2.VideoCapture(videoFilename) if capture.isOpened(): - rawCount = capture.get(cv2.cv.CV_CAP_PROP_FRAME_COUNT) + rawCount = capture.get(cv2.CAP_PROP_FRAME_COUNT) if rawCount < 0: rawCount = lastFrameNum+1 nDigits = int(floor(log10(rawCount)))+1 ret = False - capture.set(cv2.cv.CV_CAP_PROP_POS_FRAMES, firstFrameNum) + capture.set(cv2.CAP_PROP_POS_FRAMES, firstFrameNum) frameNum = firstFrameNum while frameNum<lastFrameNum and frameNum<rawCount: ret, img = capture.read() @@ -245,7 +236,7 @@ images.append(img) frameNum +=step if step > 1: - capture.set(cv2.cv.CV_CAP_PROP_POS_FRAMES, frameNum) + capture.set(cv2.CAP_PROP_POS_FRAMES, frameNum) capture.release() else: print('Video capture for {} failed'.format(videoFilename)) @@ -254,7 +245,7 @@ def getFPS(videoFilename): capture = cv2.VideoCapture(videoFilename) if capture.isOpened(): - fps = capture.get(cv2.cv.CV_CAP_PROP_FPS) + fps = capture.get(cv2.CAP_PROP_FPS) capture.release() return fps else: @@ -300,8 +291,8 @@ colorType = 'default' capture = cv2.VideoCapture(videoFilename) - width = int(capture.get(cv2.cv.CV_CAP_PROP_FRAME_WIDTH)) - height = int(capture.get(cv2.cv.CV_CAP_PROP_FRAME_HEIGHT)) + width = int(capture.get(cv2.CAP_PROP_FRAME_WIDTH)) + height = int(capture.get(cv2.CAP_PROP_FRAME_HEIGHT)) windowName = 'frame' if rescale == 1.: @@ -313,7 +304,7 @@ key = -1 ret = True frameNum = firstFrameNum - capture.set(cv2.cv.CV_CAP_PROP_POS_FRAMES, firstFrameNum) + capture.set(cv2.CAP_PROP_POS_FRAMES, firstFrameNum) if lastFrameNumArg is None: lastFrameNum = float("inf") else: @@ -374,7 +365,7 @@ cv2.imwrite('image-{{:0{}}}.png'.format(nZerosFilename).format(frameNum), img) frameNum += nFramesStep if nFramesStep > 1: - capture.set(cv2.cv.CV_CAP_PROP_POS_FRAMES, frameNum) + capture.set(cv2.CAP_PROP_POS_FRAMES, frameNum) cv2.destroyAllWindows() else: print('Cannot load file ' + videoFilename) @@ -496,11 +487,11 @@ try: flags = 0 if fixK2: - flags += cv2.cv.CV_CALIB_FIX_K2 + flags += cv2.CALIB_FIX_K2 if fixK3: - flags += cv2.cv.CV_CALIB_FIX_K3 + flags += cv2.CALIB_FIX_K3 if zeroTangent: - flags += cv2.cv.CV_CALIB_ZERO_TANGENT_DIST + flags += cv2.CALIB_ZERO_TANGENT_DIST ret, camera_matrix, dist_coeffs, rvecs, tvecs = cv2.calibrateCamera(objpoints, imgpoints, gray.shape[::-1], None, None, flags = flags) except NameError: return None
--- a/python/ml.py Fri Apr 13 16:48:02 2018 -0400 +++ b/python/ml.py Wed May 16 21:06:52 2018 -0400 @@ -25,50 +25,52 @@ ##################### def computeConfusionMatrix(model, samples, responses): - 'computes the confusion matrix of the classifier (model)' + '''computes the confusion matrix of the classifier (model) + + samples should be n samples by m variables''' classifications = {} - for x,y in zip(samples, responses): - predicted = model.predict(x) + predictions = model.predict(samples) + for predicted, y in zip(predictions, responses): classifications[(y, predicted)] = classifications.get((y, predicted), 0)+1 return classifications -class StatModel(object): - '''Abstract class for loading/saving model''' - def load(self, filename): - if path.exists(filename): - self.model.load(filename) - else: - print('Provided filename {} does not exist: model not loaded!'.format(filename)) - - def save(self, filename): - self.model.save(filename) - if opencvAvailable: - class SVM(StatModel): + class SVM(object): '''wrapper for OpenCV SimpleVectorMachine algorithm''' - def __init__(self, svmType = cv2.SVM_C_SVC, kernelType = cv2.SVM_RBF, degree = 0, gamma = 1, coef0 = 0, Cvalue = 1, nu = 0, p = 0): - self.model = cv2.SVM() - self.params = dict(svm_type = svmType, kernel_type = kernelType, degree = degree, gamma = gamma, coef0 = coef0, Cvalue = Cvalue, nu = nu, p = p) - # OpenCV3 - # self.model = cv2.SVM() - # self.model.setType(svmType) - # self.model.setKernel(kernelType) - # self.model.setDegree(degree) - # self.model.setGamma(gamma) - # self.model.setCoef0(coef0) - # self.model.setC(Cvalue) - # self.model.setNu(nu) - # self.model.setP(p) + def __init__(self, svmType = cv2.ml.SVM_C_SVC, kernelType = cv2.ml.SVM_RBF, degree = 0, gamma = 1, coef0 = 0, Cvalue = 1, nu = 0, p = 0): + self.model = cv2.ml.SVM_create() + self.model.setType(svmType) + self.model.setKernel(kernelType) + self.model.setDegree(degree) + self.model.setGamma(gamma) + self.model.setCoef0(coef0) + self.model.setC(Cvalue) + self.model.setNu(nu) + self.model.setP(p) - def train(self, samples, responses, computePerformance = False): - self.model.train(samples, responses, params = self.params) + def save(self, filename): + self.model.save(filename) + + def train(self, samples, layout, responses, computePerformance = False): + self.model.train(samples, layout, responses) if computePerformance: return computeConfusionMatrix(self, samples, responses) def predict(self, hog): - return self.model.predict(hog) + retval, predictions = self.model.predict(hog) + if hog.shape[0] == 1: + return predictions[0][0] + else: + return np.asarray(predictions, dtype = np.int).ravel().tolist() - + def SVM_load(filename): + if path.exists(filename): + svm = SVM() + svm.model = cv2.ml.SVM_load(filename) + return svm + else: + print('Provided filename {} does not exist: model not loaded!'.format(filename)) + ##################### # Clustering #####################
--- a/python/moving.py Fri Apr 13 16:48:02 2018 -0400 +++ b/python/moving.py Wed May 16 21:06:52 2018 -0400 @@ -1641,7 +1641,7 @@ croppedImg = cvutils.imageBox(img, self, instant, width, height, px, py, minNPixels) if croppedImg is not None and len(croppedImg) > 0: hog = cvutils.HOG(croppedImg, rescaleSize, orientations, pixelsPerCell, cellsPerBlock, blockNorm) - self.userTypes[instant] = int(self.appearanceClassifier.predict(hog)) + self.userTypes[instant] = self.appearanceClassifier.predict(hog.reshape(1,hog.size)) else: self.userTypes[instant] = userType2Num['unknown']
--- a/scripts/classify-objects.py Fri Apr 13 16:48:02 2018 -0400 +++ b/scripts/classify-objects.py Wed May 16 21:06:52 2018 -0400 @@ -35,10 +35,8 @@ print('Unknown speed aggregation method: {}. Exiting'.format(classifierParams.speedAggregationMethod)) sys.exit() -pedBikeCarSVM = ml.SVM() -pedBikeCarSVM.load(classifierParams.pedBikeCarSVMFilename) -bikeCarSVM = ml.SVM() -bikeCarSVM.load(classifierParams.bikeCarSVMFilename) +pedBikeCarSVM = ml.SVM_load(classifierParams.pedBikeCarSVMFilename) +bikeCarSVM = ml.SVM_load(classifierParams.bikeCarSVMFilename) # log logistic for ped and bik otherwise ((pedBeta/pedAlfa)*((sMean/pedAlfa)**(pedBeta-1)))/((1+(sMean/pedAlfa)**pedBeta)**2.) carNorm = norm(classifierParams.meanVehicleSpeed, classifierParams.stdVehicleSpeed) @@ -72,8 +70,8 @@ timeInterval.first = 0 capture = cv2.VideoCapture(videoFilename) -width = int(capture.get(cv2.cv.CV_CAP_PROP_FRAME_WIDTH)) -height = int(capture.get(cv2.cv.CV_CAP_PROP_FRAME_HEIGHT)) +width = int(capture.get(cv2.CAP_PROP_FRAME_WIDTH)) +height = int(capture.get(cv2.CAP_PROP_FRAME_HEIGHT)) #if undistort: # setup undistortion # [map1, map2] = cvutils.computeUndistortMaps(width, height, undistortedImageMultiplication, intrinsicCameraMatrix, distortionCoefficients) @@ -89,7 +87,7 @@ ret = True frameNum = timeInterval.first if not args.startFrame0: - capture.set(cv2.cv.CV_CAP_PROP_POS_FRAMES, frameNum) + capture.set(cv2.CAP_PROP_POS_FRAMES, frameNum) lastFrameNum = timeInterval.last while ret and frameNum <= lastFrameNum:
--- a/scripts/train-object-classification.py Fri Apr 13 16:48:02 2018 -0400 +++ b/scripts/train-object-classification.py Wed May 16 21:06:52 2018 -0400 @@ -2,8 +2,7 @@ import numpy as np import argparse -from cv2 import SVM_RBF, SVM_C_SVC -#from cv2.ml import SVM_RBF, SVM_C_SVC, ROW_SAMPLE # row_sample for layout in cv2.ml.SVM_load +from cv2.ml import SVM_RBF, SVM_C_SVC, ROW_SAMPLE # row_sample for layout in cv2.ml.SVM_load import cvutils, moving, ml, storage @@ -50,28 +49,28 @@ # Training the Support Vector Machine print "Training Pedestrian-Cyclist-Vehicle Model" model = ml.SVM(args.svmType, args.kernelType, args.degree, args.gamma, args.coef0, args.cvalue, args.nu, args.svmP) -classifications = model.train(np.concatenate(trainingSamplesPBV.values()), np.concatenate(trainingLabelsPBV.values()), True) +classifications = model.train(np.concatenate(trainingSamplesPBV.values()), ROW_SAMPLE, np.concatenate(trainingLabelsPBV.values()), True) if args.computeConfusionMatrix: print(classifications) model.save(args.directoryName + "/modelPBV.xml") print "Training Cyclist-Vehicle Model" model = ml.SVM(args.svmType, args.kernelType, args.degree, args.gamma, args.coef0, args.cvalue, args.nu, args.svmP) -classifications = model.train(np.concatenate(trainingSamplesBV.values()), np.concatenate(trainingLabelsBV.values()), True) +classifications = model.train(np.concatenate(trainingSamplesBV.values()), ROW_SAMPLE, np.concatenate(trainingLabelsBV.values()), True) if args.computeConfusionMatrix: print(classifications) model.save(args.directoryName + "/modelBV.xml") print "Training Pedestrian-Cyclist Model" model = ml.SVM(args.svmType, args.kernelType, args.degree, args.gamma, args.coef0, args.cvalue, args.nu, args.svmP) -classifications = model.train(np.concatenate(trainingSamplesPB.values()), np.concatenate(trainingLabelsPB.values()), True) +classifications = model.train(np.concatenate(trainingSamplesPB.values()), ROW_SAMPLE, np.concatenate(trainingLabelsPB.values()), True) if args.computeConfusionMatrix: print(classifications) model.save(args.directoryName + "/modelPB.xml") print "Training Pedestrian-Vehicle Model" model = ml.SVM(args.svmType, args.kernelType, args.degree, args.gamma, args.coef0, args.cvalue, args.nu, args.svmP) -classifications = model.train(np.concatenate(trainingSamplesPV.values()), np.concatenate(trainingLabelsPV.values()), True) +classifications = model.train(np.concatenate(trainingSamplesPV.values()), ROW_SAMPLE, np.concatenate(trainingLabelsPV.values()), True) if args.computeConfusionMatrix: print(classifications) model.save(args.directoryName + "/modelPV.xml")