def validateFace(self, dataFile, labelsFile, n, width, height, itemNo = -1):
items = readData.loadDataFile(dataFile, n, width, height)
labels = readData.loadLabelsFile(labelsFile, n)
function = 0
guess = 0
correct = 0
if itemNo != -1:
image = items[itemNo]
features = getFeatures.getFeatures(image.pixels, width, height)
for j in range(len(features)):
function += (self.weightsFace[j] * features[j])
function += self.weightsFace[len(features)]
if function >= 0: return 1
else: return 0
for i in range(len(items)):
image = items[i]
features = getFeatures.getFeatures(image.pixels, width, height)
for j in range(len(features)):
function += (self.weightsFace[j] * features[j])
function += self.weightsFace[len(features)]
if function >= 0: guess = 1
else: guess = 0
if guess == labels[i]: correct += 1
function = 0
return correct/n
def objectTracking():
numObjects = 2
maxCorners = 20
qualityLevel = 0.01
minDistance = 8
filename = "Easy.mp4"
numFrames = 2
#load video
frameData, length, h, w, fps = loadVideo(filename, 0, numFrames)
#output video
output = np.zeros([numFrames, h, w, 3])
#get bounding boxes
init_img = frameData[0, :, :, :]
init_img_gray = cv2.cvtColor(frameData[0, :, :, :], cv2.COLOR_BGR2GRAY)
bbox_list = []
bbox_pts = []
for i in range(0, numObjects):
bbox_list, bbox_pts, init_img = getBoundingBox(init_img, bbox_list,
bbox_pts)
startXs, startYs, _ = getFeatures(init_img_gray, bbox_list, maxCorners,
qualityLevel, minDistance)
output[0, :, :, :] = init_img
cnt = 0
i = 0
curr_bbox = bbox_pts
while i < length - 1:
if cnt == numFrames - 1:
frameData, _, _, _, _ = loadVideo(filename, cnt + 1, numFrames)
startXs, startYs, _ = getFeatures(init_img_gray, bbox_list,
maxCorners, qualityLevel,
minDistance)
cnt = 0
img1 = frameData[i, :, :, :]
img2 = frameData[i + 1, :, :, :]
img1_gray = cv2.cvtColor(frameData[i, :, :, :], cv2.COLOR_BGR2GRAY)
img2_gray = cv2.cvtColor(frameData[i + 1, :, :, :], cv2.COLOR_BGR2GRAY)
[newXs, newYs] = estimateAllTranslation(startXs, startYs, img)
[Xs, Ys,
new_bbox] = applyGeometricTransformation(startXs, startYs, newXs,
newYs, curr_bbox)
plotBbox(img2, new_bbox)
output[frame + 1, :, :, :] = img2
startXs = newXs
startYs = newYs
cnt += 1
def validateFace(self, dataFile, labelsFile, n, width, height, itemNo=-1):
items = readData.loadDataFile(dataFile, n, width, height)
labels = readData.loadLabelsFile(labelsFile, n)
like = 0.0
guess = 0
correct = 0
if itemNo != -1:
image = items[itemNo]
features = getFeatures.getFeatures(image.pixels, width, height)
productTrue = pow(10, 300.0)
productFalse = pow(10, 300.0)
for j in range(len(self.featurepd)):
if features[j] > 0:
productTrue *= self.featurepd[j][1]
productFalse *= self.featurepd[j][0]
else:
productTrue *= (1 - self.featurepd[j][1])
productFalse *= (1 - self.featurepd[j][0])
like = (productTrue * self.faceProb) / (productFalse *
(1 - self.faceProb))
if like >= 1:
return 1
else:
return 0
for i in range(len(items)):
image = items[i]
features = getFeatures.getFeatures(image.pixels, width, height)
productTrue = pow(10, 300.0)
productFalse = pow(10, 300.0)
for j in range(len(self.featurepd)):
if features[j] > 0:
productTrue *= self.featurepd[j][1]
productFalse *= self.featurepd[j][0]
else:
productTrue *= (1 - self.featurepd[j][1])
productFalse *= (1 - self.featurepd[j][0])
like = (productTrue * self.faceProb) / (productFalse *
(1 - self.faceProb))
if like >= 1:
guess = 1
else:
guess = 0
if guess == labels[i]:
correct += 1
return correct / n
def track_object(self):
template = cv2.cvtColor(self.img1,cv2.COLOR_BGR2GRAY)
out = cv2.VideoWriter('Results/'+self.folder_name+'_results.avi',0,cv2.VideoWriter_fourcc('M','J','P','G'),20.0,(template.shape[1],template.shape[0]))
self.startXs, self.startYs = getFeatures(template,self.bbox[0],use_shi=False)
for i in range(1,self.no_frames):
print('Tracking for frame: %s'%str(i))
self.frame1 = cv2.imread(self.folder_name+'/%s.png'%str(i-1))
self.frame2 = cv2.imread(self.folder_name+'/%s.png'%str(i))
newXs, newYs = estimateAllTranslation(self.startXs,self.startYs, self.frame1, self.frame2)
ag = applyGeometricTransformations(self.startXs, self.startYs, newXs, newYs, self.bbox[i-1])
Xs, Ys, self.bbox[i] = ag.transform()
self.center[i]= np.empty((self.no_object,1,2), dtype=int)
self.center2[i]= np.empty((self.no_object,1,2), dtype=int)
# updating coordinates
self.startXs = Xs
self.startYs = Ys
# updating feature points
no_features_left = np.sum(Xs!=-1)
print('No. of features: %d'%no_features_left)
if no_features_left<15:
print('Generate new features')
self.startXs, self.startYs = getFeatures(cv2.cvtColor(self.frame2,cv2.COLOR_BGR2GRAY),self.bbox[i])
# drawing bounding box and visualising feature point for each object
frames_draw = self.frame2.copy()
for j in range(self.no_object):
x,y,w,h = cv2.boundingRect(self.bbox[i][j,:,:].astype(int))
self.center[i][j,:,:]= np.array([x,y])
self.center2[i][j,:,:]= np.array([x,y+h])
frames_draw = cv2.rectangle(frames_draw, (x,y),(x+w,y+h), (255,0,0), 2)
for k in range(self.startXs.shape[0]):
frames_draw= cv2.circle(frames_draw, (int(self.startXs[k,j]),int(self.startYs[k,j])),3,(0,255,0),thickness=2)
center_points1=[]
center_points2=[]
for l in range(i+1):
center_points1.append(self.center[l][j,:,:])
center_points2.append(self.center2[l][j,:,:])
if(j==1):
frames_draw = cv2.drawContours(frames_draw, np.array(center_points1), -1, (0,0,255), 3)
frames_draw = cv2.drawContours(frames_draw, np.array(center_points2), -1, (0,0,255), 3)
else:
frames_draw = cv2.drawContours(frames_draw, np.array(center_points1), -1, (255,255,255), 3)
frames_draw = cv2.drawContours(frames_draw, np.array(center_points2), -1, (255,255,255), 3)
cv2.imshow("window", frames_draw)
cv2.waitKey(10)
out.write(frames_draw)
out.release()
def train(features):
train_data = None
for i in range(10):
feature, tempo = getFeatures("alda-ml/samples/" + str(i+1) + "/out.wav")
feature = np.reshape(feature[-1], -1).astype('float32')
if train_data is not None:
train_data = np.vstack([train_data, feature])
else:
train_data = np.array(feature)
# print(train_data.shape)
train_labels = None
for i in range(10):
with open("alda-ml/samples/" + str(i+1) + "/score.json") as f:
score = json.loads(f.read())
note = score['events'][0]['midi-note']
label = [0 for x in range(88)]
label[note] = 1
if train_labels is not None:
train_labels = np.vstack([train_labels, label])
else:
train_labels = np.array(label)
# print(train_labels.shape)
clf = tf.estimator.Estimator(
model_fn=cnn, model_dir="tmp/coolModel")
tensorTrain(train_data, train_labels, clf)
return clf
def trainFace(self, dataFile, labelsFile, n, width, height, randomList):
items = readData.loadDataFile(dataFile, n, width, height)
labels = readData.loadLabelsFile(labelsFile, n)
self.featurepd = []
numberFaces = 0
# featureTimes is a list of tuples where the first number is the number of times the feature occurs when the image is not a face
# and the second number is the number of times the features occurs when the image is a face
featureTimes = []
numberFeatures = (math.ceil(height / getFeatures.DIM)) * math.ceil(
(width / getFeatures.DIM))
for i in range(numberFeatures):
featureTimes.append([0.0, 0.0])
for count in randomList:
image = items[count]
if labels[count] == 1:
numberFaces += 1.0
features = getFeatures.getFeatures(image.pixels, width, height)
for i in range(len(features)):
if features[i] > 0:
if labels[count] == 1:
featureTimes[i][1] += 1
else:
featureTimes[i][0] += 1
for i in range(len(featureTimes)):
self.featurepd.append([
(featureTimes[i][0] + self.k) /
(len(randomList) - numberFaces + self.k),
(featureTimes[i][1] + self.k) / (numberFaces + self.k)
])
self.faceProb = numberFaces / len(randomList)
def main():
# --------------------------------------Get Preprocessed Data------------------------------------------------------
a = GetDirs()
f1 = a.getDirs('DataFolder/CGMSeriesLunch')
f2 = a.getDirs('DataFolder/CGMTimeSeries')
zippedTups = a.zipFiles(f1, f2)
allArraysLunch = np.array([])
allArraysdateNum = np.array([])
allArraysLunch = np.concatenate(
[a.returnArrays(i[0], i[1], 'Lunch') for i in zippedTups])
allArraysdateNum = np.concatenate(
[a.returnArrays(i[0], i[1]) for i in zippedTups])
# ------------------------------------------Get Features-----------------------------------------------------------
getF = getFeatures(allArraysLunch)
F1 = getF.fft()
F2 = getF.entropy()
F3 = getF.skewness()
F4 = getF.movingStd()
# ----------------------------------------------Get PCA------------------------------------------------------------
PCs = getPCA(np.concatenate((F1, F2[:, None], F3[:, None], F4),
axis=1)).pca()
# -----------------------------Get PCA Results in decreasing order of accuracy--------------------------------------
# print(getPCA(np.concatenate((F1, F2[:, None], F3[:, None], F4), axis=1)).results())
# --------------------------------------------Plot PCA--------------------------------------------------------------
plotDiag("Principal Conponent Analysis", 'Red', PCs['PC1'], 'CGM Data',
'PCA Values', 'PC1').plot()
def load_letter(folder):
"""Load the data for a single letter label."""
image_files = os.listdir(folder)
dataset = np.ndarray(shape=(len(image_files), 41), dtype=np.float32)
print(folder)
num_images = 0
for image in image_files:
image_file = os.path.join(folder, image)
try:
im = cv2.imread(image_file)
if im.shape != (image_size, image_size, 3):
raise Exception('%s\nUnexpected image shape: %s' %
(image_file, str(im.shape)))
features = getFeatures(im)
dataset[num_images, :] = features
num_images = num_images + 1
except IOError as e:
print('Could not read:', image_file, ':', e,
'- it\'s ok, skipping.')
dataset = dataset[:num_images, :]
print('Full dataset tensor:', dataset.shape)
print('Mean:', np.mean(dataset))
print('Standard deviation:', np.std(dataset))
return dataset
def addNewClass(jsonGMM, newClassName, newClassData=None):
"""
Incorporate a new class into an existing GMM JSON model (i.e. list not dict!)
@param jsonGMM: List containing GMMs, see createJSON.py for exact structure
@param newClassName: Name of the new class. Has to match the name of the folder where the sound files are located
@param newClassData: Already extracted MFCC Features for the given newClassName. If not provided, a feature extraction will be performed
@return: List containing the added class as the last element
"""
jGMM = copy.deepcopy(jsonGMM)
newClassData = getFeatures(newClassName)
scaler = preprocessing.StandardScaler()
scaler.mean_ = jGMM[0]["scale_means"]
scaler.std_ = jGMM[0]["scale_stddevs"]
X_train = scaler.transform(newClassData)
n_train_new = X_train.shape[0]
n_features = X_train.shape[1]
n_components = 16
print("Start training new class " + newClassName)
newClf = GMM(n_components = n_components, covariance_type='full', n_iter=100) # set n_iter to 100 to speed up the training
newClf.fit(X_train)
""" Update the dict containing mapping of class names: """
newClassDict = dict(jGMM[0]['classesDict'])
newClassDict[newClassName] = len(newClassDict.values())
new_n_classes = jGMM[0]["n_classes"] + 1
""" Update classesDict and n_classes for all old classes: """
for i in range(len(jGMM)):
jGMM[i]["classesDict"] = newClassDict
jGMM[i]["n_classes"] = new_n_classes
""" Add the new class to the list: """
jGMM.append({})
jGMM[-1]["classesDict"] = newClassDict
jGMM[-1]["n_classes"] = new_n_classes
jGMM[-1]["scale_mean"] = scaler.mean_
jGMM[-1]["scale_stddev"] = scaler.std_
jGMM[-1]["n_components"] = n_components
jGMM[-1]["n_features"] = n_features
jGMM[-1]["n_train"] = n_train_new
jGMM[-1]["weights"] = newClf.weights_.tolist()
jGMM[-1]["means"] = newClf.means_.tolist()
jGMM[-1]["covars"] = newClf.covars_.tolist()
return jGMM
def train():
features=[]
for i in range(1,35):
if i==6 or i==8:
continue
filename='SSM'+str(i)
features.append(GF.getFeatures(filename))
print i
numpyfeatures=np.array(features)
np.savetxt('TrainingData', numpyfeatures)
print numpyfeatures.shape
def train():
features = []
for i in range(1, 35):
if i == 6 or i == 8:
continue
filename = 'SSM' + str(i)
features.append(GF.getFeatures(filename))
print i
numpyfeatures = np.array(features)
np.savetxt('TrainingData', numpyfeatures)
print numpyfeatures.shape
def getDataSet():
X, Y = getFeatures()
features = len(X[0])
cases = len(X)
DS = SupervisedDataSet(features, 1)
i = 0
while (i < cases):
DS.addSample(X[i], Y[i])
i += 1
TrainDS, TestDS = DS.splitWithProportion(0.7)
return TrainDS, TestDS
def BasicPCAGraphs(vecnumber):
countyvecs, diagvecs, diagdict = getFeatures.getFeatures()
pca = PCA(n_components=3)
standardscaler = preprocessing.StandardScaler()
veclist = getVecList(countyvecs, vecnumber, "Medicare")
vecs = standardscaler.fit_transform(veclist)
Medicareprincipalcomponents = pca.fit_transform(vecs)
Medicarevar = pca.explained_variance_ratio_
x1, y1, z1 = Medicareprincipalcomponents.T
veclist = getVecList(countyvecs, vecnumber, "Medicaid")
vecs = standardscaler.fit_transform(veclist)
Medicaidprincipalcomponents = pca.fit_transform(vecs)
Medicaidvar = pca.explained_variance_ratio_
x2, y2, z2 = Medicaidprincipalcomponents.T
veclist = getVecList(countyvecs, vecnumber, "Uninsured")
vecs = standardscaler.fit_transform(veclist)
Uninsuredprincipalcomponents = pca.fit_transform(vecs)
Uninsuredvar = pca.explained_variance_ratio_
x3, y3, z3 = Uninsuredprincipalcomponents.T
plt.plot(x1, y1, 'ro', x2, y2, 'bo', x3, y3, 'go')
plt.title("PCA of Procedure Based Vectors, Colored By Payment Method")
plt.show()
diagveclist = getVecList(diagvecs, -1, "Medicare")
vecs = standardscaler.fit_transform(diagveclist)
Medicareprincipalcomponents = pca.fit_transform(vecs)
Medicarevar = pca.explained_variance_ratio_
x1, y1, z1 = Medicareprincipalcomponents.T
diagveclist = getVecList(diagvecs, -1, "Medicaid")
vecs = standardscaler.fit_transform(diagveclist)
Medicaidprincipalcomponents = pca.fit_transform(vecs)
Medicaidvar = pca.explained_variance_ratio_
x2, y2, z2 = Medicaidprincipalcomponents.T
diagveclist = getVecList(diagvecs, -1, "Uninsured")
vecs = standardscaler.fit_transform(diagveclist)
Uninsuredprincipalcomponents = pca.fit_transform(vecs)
Uninsuredvar = pca.explained_variance_ratio_
x3, y3, z3 = Uninsuredprincipalcomponents.T
plt.plot(x1, y1, 'ro', x2, y2, 'bo', x3, y3, 'go')
plt.title("PCA of Diagnosis Based Vectors, Colored By Payment Method")
plt.show()
def displayFeatures(filename):
numObjects = 2
maxCorners = 20
qualityLevel = 0.01
minDistance = 8
im = cv2.imread(filename)
im = cv2.cvtColor(im, cv2.COLOR_BGR2GRAY)
bbox_list = []
bbox_pts = []
for i in range(0, numObjects):
bbox_list, bbox_pts, newImage = getBoundingBox(im, bbox_list, bbox_pts)
print(bbox_pts)
x, y, _ = getFeatures(im, bbox_list, maxCorners, qualityLevel, minDistance)
print(x)
print(y)
def test_detect(vidfname, outfname):
#vid1 = imageio.get_reader("shifted.mp4", 'ffmpeg')
vid1 = imageio.get_reader(vidfname, 'ffmpeg')
prevframe = None
prevbboxes = None
prevXs, prevYs = np.zeros((1, 1)), np.zeros((1, 1))
allimg = []
for i, frame in enumerate(vid1):
print i
if i % 30 == 0 or count_minfeats(prevXs) < 20:
print "RECALCULATING"
try:
newbboxes = detectFace(frame)
except AttributeError:
print "NO NEW FACE FOUND! Trying again on next frame"
continue
greyframe = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
Xs, Ys = getFeatures(greyframe, newbboxes)
print Xs.shape
prevXs, prevYs = Xs, Ys
else:
Xs, Ys = estimateAllTranslation(prevXs, prevYs, prevframe, frame)
Xs, Ys, newbboxes = applyGeometricTransformation(
prevXs, prevYs, Xs, Ys, prevbboxes)
#plot_all(frame, newbboxes, Xs, Ys, display=True)
allimg.append(plot_all(frame, newbboxes, Xs, Ys, display=False))
# if i==0 or (i > 9 and not i % 10):
# plot_features(frame, Xs, Ys, display=True)
# print Xs-prevXs
prevbboxes = newbboxes
prevframe = frame
prevXs, prevYs = Xs.copy(), Ys.copy()
# if i > 60:
# print "Quitting early for testing purposes"
# break
imageio.mimsave(outfname, allimg)
return
def trainFace(self, dataFile, labelsFile, n, width, height, rList):
items = readData.loadDataFile(dataFile, n, width, height)
old_labels = readData.loadLabelsFile(labelsFile, n)
self.weightsFace = []
numberFeatures = (math.ceil(height/getFeatures.DIM)) * math.ceil((width/getFeatures.DIM))
for i in range(numberFeatures + 1):
self.weightsFace.append(0)
labels = []
for r in rList:
labels.append(old_labels[r])
# this array contains the features for every image
features = []
for i in rList:
features.append(getFeatures.getFeatures(items[i].pixels, width, height))
loop = 0
index = 0
# traverses all the items and updates the weights accordingly
# stops when a full loop is completed without updates
iterations = 0
while loop < len(rList) and iterations < 10 *len(rList):
iterations +=1
function = 0
for i in range(numberFeatures):
function += self.weightsFace[i] * features[index][i]
function += self.weightsFace[numberFeatures]
if (function >= 0 and labels[index] == 1) or (function < 0 and labels[index] == 0):
loop += 1
elif function < 0 and labels[index] == 1:
for i in range(numberFeatures):
self.weightsFace[i] += features[index][i]
self.weightsFace[numberFeatures] += 1
loop = 0
else:
for i in range(numberFeatures):
self.weightsFace[i] -= features[index][i]
self.weightsFace[numberFeatures] -= 1
loop = 0
if index == len(rList)-1: index = 0
else: index += 1
def learnFeatureExistance(busImportantFeatures, userImportantFeatures, trainReviews, path):
logger = logging.getLogger('signature.lFE.learnFE')
logger.info('starting learnFeatureExistance from %d reviews'%len(trainReviews))
fsw = featureStructureWorker()
modelDict = dict()
trainAveragesDict = dict()
for f, feature in enumerate(fsw.featureIdicator):
if not fsw.featureIdicator[feature]:
continue
logger.debug('Start working with %s'%feature)
#get data
X, Y, trainAveragesDict[feature] = getFeatures(logger, feature, trainReviews, busImportantFeatures, userImportantFeatures,
trainAverages = {}, is_train = True)
logger.debug('Got features for %d reviews'%len(X))
# #cross validation
# indicator = range(len(X))
# random.shuffle(indicator)
# thres = int(len(indicator)*0.8)
# trainX = np.array([X[i] for i in indicator[:thres]])
# trainY = np.array([Y[i] for i in indicator[:thres]])
# testX = np.array([X[i] for i in indicator[thres:]])
# testY = np.array([Y[i] for i in indicator[thres:]])
#Logistic Regression
#bestThres,bestF1,logmodel = getBestLogModel(logger, feature, trainX, trainY, testX, testY, X, Y, path)
bestThres,bestF1,logmodel = getBestLogModel(logger, feature, X, Y, path)
#bestThresSVM,bestF1SVM,svmmodel = getBestSVMModel(logger, feature, X, Y, path)
# crossValidation(logger, np.array(X), np.array(Y))
modelDict[feature] = [bestThres,bestF1,logmodel]
# print f
# if f > 6:
# break
return trainAveragesDict, modelDict
def learnTopicExistence(busImportantFeatures, userImportantFeatures, trainReviews, path):
logger = logging.getLogger('signature.lTE.learnTopicExistence')
logger.info('starting learnTopicExistence from %d reviews'%len(trainReviews))
fsw = featureStructureWorker()
modelDict = dict()
for f, topic in enumerate(fsw.featureIdicator):
if not fsw.featureIdicator[topic]:
continue
logger.debug('Start working with %s'%topic)
#get data
X, Y = getFeatures(logger, topic, trainReviews, busImportantFeatures, userImportantFeatures,
trainAverages = {}, is_train = True)
logger.debug('Got features for %d reviews'%len(X))
modelDict[topic] = getBestModel(logger, topic, X, Y, path)
#print modelDict
return modelDict
def objectTracking(rawVideo,
draw_bb=False,
play_realtime=False,
save_to_file=False):
# initilize
n_frame = 400
frames = np.empty((n_frame, ), dtype=np.ndarray)
frames_draw = np.empty((n_frame, ), dtype=np.ndarray)
bboxs = np.empty((n_frame, ), dtype=np.ndarray)
for frame_idx in range(n_frame):
_, frames[frame_idx] = rawVideo.read()
# draw rectangle roi for target objects, or use default objects initilization
if draw_bb:
n_object = int(input("Number of objects to track:"))
bboxs[0] = np.empty((n_object, 4, 2), dtype=float)
for i in range(n_object):
(xmin, ymin, boxw, boxh) = cv2.selectROI("Select Object %d" % (i),
frames[0])
cv2.destroyWindow("Select Object %d" % (i))
bboxs[0][i, :, :] = np.array([[xmin, ymin], [xmin + boxw, ymin],
[xmin, ymin + boxh],
[xmin + boxw,
ymin + boxh]]).astype(float)
else:
n_object = 1
bboxs[0] = np.array([[[291, 187], [405, 187], [291, 267],
[405, 267]]]).astype(float)
if save_to_file:
out = cv2.VideoWriter('output.avi', 0,
cv2.VideoWriter_fourcc('M', 'J', 'P', 'G'), 20.0,
(frames[i].shape[1], frames[i].shape[0]))
# Start from the first frame, do optical flow for every two consecutive frames.
startXs, startYs = getFeatures(cv2.cvtColor(frames[0], cv2.COLOR_RGB2GRAY),
bboxs[0],
use_shi=False)
for i in range(1, n_frame):
print('Processing Frame', i)
newXs, newYs = estimateAllTranslation(startXs, startYs, frames[i - 1],
frames[i])
Xs, Ys, bboxs[i] = applyGeometricTransformation(
startXs, startYs, newXs, newYs, bboxs[i - 1])
# update coordinates
startXs = Xs
startYs = Ys
# update feature points as required
n_features_left = np.sum(Xs != -1)
print('# of Features: %d' % n_features_left)
if n_features_left < 15:
print('Generate New Features')
startXs, startYs = getFeatures(
cv2.cvtColor(frames[i], cv2.COLOR_RGB2GRAY), bboxs[i])
# draw bounding box and visualize feature point for each object
frames_draw[i] = frames[i].copy()
for j in range(n_object):
(xmin, ymin, boxw,
boxh) = cv2.boundingRect(bboxs[i][j, :, :].astype(int))
frames_draw[i] = cv2.rectangle(frames_draw[i], (xmin, ymin),
(xmin + boxw, ymin + boxh),
(255, 0, 0), 2)
for k in range(startXs.shape[0]):
frames_draw[i] = cv2.circle(
frames_draw[i], (int(startXs[k, j]), int(startYs[k, j])),
3, (0, 0, 255),
thickness=2)
# imshow if to play the result in real time
if play_realtime:
cv2.imshow("win", frames_draw[i])
cv2.waitKey(10)
if save_to_file:
out.write(frames_draw[i])
if save_to_file:
out.release()
def applyFeatureExistance(busImportantFeatures, userImportantFeatures, testReviews, modelDict, trainAveragesDict, path):
logger = logging.getLogger('signature.aFE.applyFE')
logger.info('starting applyFeatureExistance from %d reviews'%len(testReviews))
fsw = featureStructureWorker()
featureWeights = dict()
featureF1 = dict()
for i, feature in enumerate(fsw.featureIdicator):
if not fsw.featureIdicator[feature]:
continue
logger.debug('Start working with %s'%feature)
#get data
X, Y = getFeatures(logger, feature, testReviews, busImportantFeatures, userImportantFeatures,
trainAverages = trainAveragesDict[feature], is_train = False)
#weight = frequency
featureWeights[feature] = float(list(Y).count(1))/len(Y)
Ypred = [x[1] for x in modelDict[feature][2].predict_proba(np.array(X))]
Yreal = np.array(Y)
Ybus = []
for review in testReviews:
busID = review['business_id']
if busID in busImportantFeatures:
pfreq = busImportantFeatures[busID]['featureFreq'].get(feature,0.0)
else:
pfreq = featureWeights[feature]
Ybus.append(pfreq)
featureF1[feature] = drawPR(feature,Yreal,Ypred,Ybus, modelDict[feature][0], path)
for r, review in enumerate(testReviews):
#reviewFeatures = fsw.getReviewFeaturesExistence(review['features'])
review['exPredFeatures'] = review.get('exPredFeatures', {})
existence = Yreal[r]
#print Yreal[r], Ypred[r], modelDict[feature][0]
if Ypred[r] >= modelDict[feature][0]:
predictedExistence = 1
else:
predictedExistence = 0
#check if feature important
if existence + predictedExistence > 0.5:
review['exPredFeatures'][feature] = [existence, predictedExistence]
#print review['exPredFeatures']
if not r%10000:
logger.debug('%d reviews processed'%r)
Jaccard = list()
Jaccard_weighted = list()
Jaccard_baseline = list()
Jaccard_baseline_weighted = list()
TP = 0
FP = 0
FN = 0
TP_all = 0
FP_all = 0
FN_all = 0
TP_bus = 0
FP_bus = 0
FN_bus = 0
TP_int = 0
FP_int = 0
FN_int = 0
for r, review in enumerate(testReviews):
Jaccard_intersection = 0.0
Jaccard_union = 0.0
Jaccard_intersection_weighted = 0.0
Jaccard_union_weighted = 0.0
Jaccard_intersection_baseline = 0.0
Jaccard_union_baseline = 0.0
Jaccard_intersection_baseline_weighted = 0.0
Jaccard_union_baseline_weighted = 0.0
busID = review['business_id']
if busID in busImportantFeatures:
busAspects = set([f for f in busImportantFeatures[busID]['featureFreq'] if busImportantFeatures[busID]['featureFreq'][f] > 10 and
busImportantFeatures[busID]['sentiment'][f][1] > 1])
else:
busAspects = set([f for f in fsw.featureIdicator if fsw.featureIdicator[feature]])
# userID = review['user_id']
# if userID in userImportantFeatures:
# userAspects = set([f for f in userImportantFeatures[userID]['featureFreq'] if userImportantFeatures[userID]['featureFreq'][f] > 10 and
# userImportantFeatures[userID]['sentiment'][f][1] > 1])
# else:
# userAspects = set([f for f in fsw.featureIdicator if fsw.featureIdicator[feature]])
#interBU = userAspects.intersection(busAspects)
#buildin INTERSECTION
busID = review['business_id']
if busID in busImportantFeatures:
busImpAspects = set([f for f in busImportantFeatures[busID]['featureFreq'] if busImportantFeatures[busID]['featureFreq'][f] > 50 and
busImportantFeatures[busID]['sentiment'][f][1] > 1])
busIntAspects = set([f for f in busImportantFeatures[busID]['featureFreq'] if busImportantFeatures[busID]['featureFreq'][f] > 10 and
busImportantFeatures[busID]['sentiment'][f][1] > 1])
else:
busImpAspects = set([f for f in fsw.featureIdicator if fsw.featureIdicator[feature]])
busIntAspects = set([f for f in fsw.featureIdicator if fsw.featureIdicator[feature]])
userID = review['user_id']
if userID in userImportantFeatures:
userAspects = set([f for f in userImportantFeatures[userID]['featureFreq'] if userImportantFeatures[userID]['featureFreq'][f] > 10 and
userImportantFeatures[userID]['sentiment'][f][1] > 1])
else:
userAspects = set([f for f in fsw.featureIdicator if fsw.featureIdicator[feature]])
interBU = busImpAspects.union(userAspects.intersection(busIntAspects))
for feature in review['exPredFeatures']:
if review['exPredFeatures'][feature] == [1,1]:
TP += 1
elif review['exPredFeatures'][feature] == [0,1]:
FP += 1
if review['exPredFeatures'][feature] == [1,0]:
FN += 1
#baseline all
if review['exPredFeatures'][feature][0] == 1:
TP_all += 1
#baseline business
if feature in busAspects and review['exPredFeatures'][feature][0] == 1:
TP_bus += 1
elif feature in busAspects and review['exPredFeatures'][feature][0] == 0:
FP_bus += 1
elif feature not in busAspects and review['exPredFeatures'][feature][0] == 1:
FN_bus += 1
#baseline intersection
if feature in interBU and review['exPredFeatures'][feature][0] == 1:
TP_int += 1
elif feature in interBU and review['exPredFeatures'][feature][0] == 0:
FP_int += 1
elif feature not in interBU and review['exPredFeatures'][feature][0] == 1:
FN_int += 1
#print TP_int, FP_int, FN_int
if review['exPredFeatures'][feature] == [1,1]:
Jaccard_intersection += 1.0
Jaccard_intersection_weighted += featureWeights[feature]
Jaccard_union += 1.0
Jaccard_union_weighted += featureWeights[feature]
if review['exPredFeatures'][feature][0] == 1:
Jaccard_intersection_baseline += 1.0
Jaccard_intersection_baseline_weighted += featureWeights[feature]
for feature in fsw.featureIdicator:
if fsw.featureIdicator[feature]:
FP_all += 1
Jaccard_union_baseline += 1
Jaccard_union_baseline_weighted += featureWeights[feature]
if Jaccard_union:
Jaccard.append(Jaccard_intersection/Jaccard_union)
if Jaccard_union_weighted:
Jaccard_weighted.append(Jaccard_intersection_weighted/Jaccard_union_weighted)
if Jaccard_union_baseline:
Jaccard_baseline.append(Jaccard_intersection_baseline/Jaccard_union_baseline)
if Jaccard_union_baseline_weighted:
Jaccard_baseline_weighted.append(Jaccard_intersection_baseline_weighted/Jaccard_union_baseline_weighted)
#SIGNATURE METHOD
Presision = float(TP)/(TP+FP)
Recall = float(TP)/(TP+FN)
F1 = 2*Presision*Recall/(Presision+Recall)
PreRec = [Presision,Recall,F1]
#baseline ALL
Presision_all = float(TP_all)/(TP_all+FP_all)
Recall_all = float(TP_all)/(TP_all+FN_all)
F1_all = 2*Presision_all*Recall_all/(Presision_all+Recall_all)
PreRec_all = [Presision_all,Recall_all,F1_all]
#baseline BUSINESS
Presision_bus = float(TP_bus)/(TP_bus+FP_bus)
Recall_bus = float(TP_bus)/(TP_bus+FN_bus)
F1_bus = 2*Presision_bus*Recall_bus/(Presision_bus+Recall_bus)
PreRec_bus = [Presision_bus,Recall_bus,F1_bus]
#print TP_int, FP_int
#baseline INTERSECTION
Presision_int = float(TP_int)/(TP_int+FP_int)
Recall_int = float(TP_int)/(TP_int+FN_int)
F1_int = 2*Presision_int*Recall_int/(Presision_int+Recall_int)
PreRec_int = [Presision_int,Recall_int,F1_int]
return testReviews, featureWeights, [[np.average(Jaccard), np.average(Jaccard_weighted)],
[np.average(Jaccard_baseline),
np.average(Jaccard_baseline_weighted)]], featureF1, [PreRec,PreRec_all,
PreRec_bus, PreRec_int]
def faceTracking(rawVideo):
cap = cv2.VideoCapture(rawVideo)
output = None
pre_img = None
# first frame
ret, cur_img = cap.read()
bbox = detectFace(cur_img)
startXs, startYs = getFeatures(cur_img, bbox)
# initialize video writer
h, w, l = cur_img.shape
fourcc = cv2.VideoWriter_fourcc('m', 'p', '4', 'v')
# change tracked_video name for each run
tracked_video = './Output_Video/tracked_video.avi'
output = cv2.VideoWriter(tracked_video, fourcc, 20, (w, h), True)
# draw box on first frame
imgwbox = drawBox(cur_img, bbox)
output.write(imgwbox)
pre_img = cur_img
count = 0
while (cap.isOpened()):
ret, cur_img = cap.read()
if not ret:
break
newXs, newYs = estimateAllTranslation(startXs, startYs, pre_img,
cur_img)
Xs, Ys, newbbox = applyGeometricTransformation(startXs, startYs, newXs,
newYs, bbox)
box_features = np.array([])
for i in range(len(Xs)):
box_features = np.append(box_features, len(Xs[i]))
print sum(box_features)
if sum(box_features) < 10:
newbbox = detectFace(cur_img)
Xs, Ys = getFeatures(cur_img, newbbox)
pre_img = cur_img
startXs = Xs
startYs = Ys
bbox = newbbox
imgwbox = drawBox(cur_img, bbox)
output.write(imgwbox)
# print video record
print('{} frame finished').format(count)
count += 1
# close video writer
cv2.destroyAllWindows()
cap.release()
output.release()
return tracked_video
currVal=[]
#print str(len(item))
#print
#if d >=int(start) and d< int(end):
if d>=0:
#print item
print
soup = BeautifulSoup(item)
script = soup.find("script")
ret ={}
#condition for top level script
if script is not None:
if script.parent.name == "head" :
actItem = script.text
ret = getFeatures.getFeatures(cleanString(actItem))
else:
ret = getFeatures.getFeatures(cleanString(item))
else:
ret = getFeatures.getFeatures(cleanString(item))
#print ".....................................................SCRIPT COUNT" + str(d+1)
#skip this entry if could not be parsed
if len(ret.keys())!=0:
valid+=1
for key in ret.keys():
currVal.append(ret[key])
#print "current key is " + str(key) + " and value is " + str(ret[key])
#append a 1 for malicious
currVal.append("0")
def faceTracking(rawVideo):
#TODO: Your code here
import numpy as np
import cv2
import scipy
import matplotlib.pyplot as plt
import matplotlib
from detectFace import detectFace
from getFeatures import getFeatures
from estimateAllTranslation import estimateAllTranslation
from applyGeometricTransformation import applyGeometricTransformation
frameSet = [] #list to store all frames of the input video
newFrameSet = [] #list to store all frames of the output video
tf = True
plt.ioff()
while tf: #read each frame in the input video
tf, frame = rawVideo.read()
frameSet.append(frame)
frameSet = frameSet[:-1]
bbox = detectFace(frameSet[0]) #detect bounding box in the first frame
gray = cv2.cvtColor(
frameSet[0],
cv2.COLOR_BGR2GRAY) #convert first frame to gray scale image
x, y = getFeatures(
gray,
bbox) #extract feature points from gray scale image and bounding box
#drawing
plt.imshow(cv2.cvtColor(frameSet[0], cv2.COLOR_BGR2RGB))
[r1b, c1b, d1b] = np.asarray(bbox.shape)
for i in range(r1b): #plot bounding box and feature points
b = bbox[i, :, :]
xloc = x[:, i]
yloc = y[:, i]
facebb = matplotlib.patches.Polygon(b, closed=True, fill=False)
# facebb.set_edgecolor('w')#uncomment if bounding box color blends with black background
features = plt.plot(xloc, yloc, 'w.', ms=1)
plt.gca().add_patch(facebb)
plt.axis('off')
plt.savefig("temp.png", dpi=300, bbox_inches="tight")
img = cv2.imread("temp.png")
plt.close()
newFrameSet.append(img)
#getting features and transforming
for k in range(
1, len(frameSet)): #iterate through all frames of the input video
newXs, newYs = estimateAllTranslation(x, y, frameSet[k - 1],
frameSet[k])
Xs, Ys, newbbox = applyGeometricTransformation(x, y, newXs, newYs,
bbox)
plt.imshow(cv2.cvtColor(frameSet[k], cv2.COLOR_BGR2RGB))
print len(Xs)
for j in range(r1b):
b = newbbox[j, :, :]
xloc = Xs[:, j]
yloc = Ys[:, j]
facebb = matplotlib.patches.Polygon(b, closed=True, fill=False)
# facebb.set_edgecolor('w')#uncomment if bounding box color blends with black background
features = plt.plot(xloc, yloc, 'w.', ms=1)
plt.gca().add_patch(facebb)
plt.axis('off')
plt.savefig("temp.png", dpi=300, bbox_inches="tight")
img = cv2.imread("temp.png")
plt.close()
newFrameSet.append(img)
x = Xs
y = Ys
bbox = newbbox
[height, width, layer] = np.asarray(newFrameSet[0].shape)
trackedVideo = cv2.VideoWriter(
'output.mp4', cv2.VideoWriter_fourcc(*'MP4V'), 30,
(width, height)) #write frames to a video in mp4 format and 30fps
for m in range(len(newFrameSet)):
trackedVideo.write(newFrameSet[m].astype('uint8'))
cv2.destroyAllWindows()
trackedVideo.release()
return trackedVideo
def begin(filename):
illumination_correction.illu_Correct(filename)
Otsu_Segmentation.segment(filename)
getFeatures.getFeatures(filename)
def objectTracking(rawVideo):
video_arr = get_video_as_numpy(rawVideo)
img1 = video_arr[0]
img1_grey = img1.dot([0.299, 0.587, 0.114])
h, w = img1_grey.shape
raw_bbox, class_names = mrcnn_detect(img1)
bbox = np.zeros((len(raw_bbox), 4, 2), dtype=int)
for k in range(len(raw_bbox)):
x_low, y_low, x_high, y_high = raw_bbox[k]
bbox[k, 0, :] = (x_low, y_low)
bbox[k, 1, :] = (x_high, y_low)
bbox[k, 2, :] = (x_high, y_high)
bbox[k, 3, :] = (x_low, y_high)
# params for ShiTomasi corner detection
feature_params = dict(maxCorners=500,
qualityLevel=0.0001,
minDistance=3,
blockSize=3)
# Parameters for lucas kanade optical flowimg1_grey
lk_params = dict(winSize=(15, 15),
maxLevel=2,
criteria=(cv2.TERM_CRITERIA_EPS | cv2.TERM_CRITERIA_COUNT,
10, 0.03))
p0 = cv2.goodFeaturesToTrack(img1_grey.astype('uint8'),
mask=None,
**feature_params)
bg_mask = bgdMask(bbox, img1_grey)
plt.figure()
plt.imshow(bg_mask)
plt.show()
pxs = np.clip(p0[:, :, 1], 0, h - 1).astype(int)
pys = np.clip(p0[:, :, 0], 0, w - 1).astype(int)
is_bg = bg_mask[pxs, pys]
bg_p0 = p0[is_bg].reshape(-1, 1, 2)
bbox_h_threshold = abs(bbox[0][1][0] - bbox[0][0][0]) * 2.5
bbox_w_threshold = abs(bbox[0][3][1] - bbox[0][0][1]) * 2.5
startXs, startYs = getFeatures(img1_grey, bbox)
starting_num_features = startXs.shape[0]
oldXs, oldYs = startXs, startYs
oldFrame = img1.copy()
for obj in bbox:
x1, y1 = int(np.round(obj[0][1])), int(np.round(obj[0][0]))
x2, y2 = int(np.round(obj[2][1])), int(np.round(obj[2][0]))
cv2.rectangle(img1, (x1, y1), (x2, y2), (255, 150, 150), 1)
n, f = startXs.shape
for i in range(n):
for j in range(f):
if startXs[i][j] == -1:
continue
newVideoName = 'output_videos/' + rawVideo.split('/')[1].split(
'.')[-2] + '_result' + '.avi'
_, h, w, _ = video_arr.shape
writer = cv2.VideoWriter(newVideoName,
cv2.VideoWriter_fourcc('M', 'J', 'P', 'G'),
25, (w, h),
isColor=True)
writer.write(img1[:, :, [2, 1, 0]])
traj_x = []
traj_y = []
for idx in range(1, len(video_arr)):
newFrame = video_arr[idx]
newXs, newYs = estimateAllTranslation(oldXs, oldYs, oldFrame, newFrame)
new_grey = newFrame.dot([0.299, 0.587, 0.114])
old_gray = oldFrame.dot([0.299, 0.587, 0.114])
bg_p1, st, err = cv2.calcOpticalFlowPyrLK(old_gray.astype('uint8'),
new_grey.astype('uint8'),
bg_p0, None, **lk_params)
restXs, restYs, newbbox = applyGeometricTransformation(
oldXs, oldYs, newXs, newYs, bbox)
if len(bg_p0) == 0:
break
movingBox = IsMoving(bg_p0[st == 1], bg_p1[st == 1], bbox, newbbox,
0.5, new_grey)
traj_x += list(restXs.flatten())
traj_y += list(restYs.flatten())
if len(restXs) < 2:
break
bg_mask = bgdMask(newbbox, new_grey)
p0 = cv2.goodFeaturesToTrack(new_grey.astype('uint8'),
mask=None,
**feature_params)
try:
pxs = np.clip(p0[:, :, 1], 0, h - 1).astype(int)
pys = np.clip(p0[:, :, 0], 0, w - 1).astype(int)
is_bg = bg_mask[pxs, pys]
bg_p0 = p0[is_bg].reshape(-1, 1, 2)
except:
break
oldFrame = newFrame.copy()
oldXs, oldYs = restXs, restYs
bbox = newbbox.copy()
old_class_names = class_names
if idx % 50 == 0:
raw_bbox, class_names = mrcnn_detect(video_arr[idx])
bbox = np.zeros((len(raw_bbox), 4, 2), dtype=int)
for k in range(len(raw_bbox)):
x_low, y_low, x_high, y_high = raw_bbox[k]
bbox[k, 0, :] = (x_low, y_low)
bbox[k, 1, :] = (x_high, y_low)
bbox[k, 2, :] = (x_high, y_high)
bbox[k, 3, :] = (x_low, y_high)
oldXs, oldYs = getFeatures(new_grey, bbox)
for i in range(len(newbbox)):
obj = newbbox[i]
if np.isnan(obj).any():
continue
if outOfBounds(obj, new_grey):
continue
x1, y1 = int(np.round(obj[0][1])), int(np.round(obj[0][0]))
x2, y2 = int(np.round(obj[2][1])), int(np.round(obj[2][0]))
font = cv2.FONT_HERSHEY_SIMPLEX
bottomLeftCornerOfText = ((x1, y1))
fontScale = 0.6
lineType = 2
if movingBox[i]:
cv2.rectangle(newFrame, (x1, y1), (x2, y2), (255, 150, 150), 2)
cv2.putText(newFrame, 'moving ' + old_class_names[i],
bottomLeftCornerOfText, font, fontScale,
(255, 150, 150), lineType)
else:
cv2.rectangle(newFrame, (x1, y1), (x2, y2), (0, 150, 150), 2)
cv2.putText(newFrame, 'still ' + old_class_names[i],
bottomLeftCornerOfText, font, fontScale,
(0, 150, 150), lineType)
n, f = oldXs.shape
writer.write(newFrame[:, :, [2, 1, 0]])
return newVideoName
def faceTracking(rawVideo):
#process the video here, convert into frames of images
#assuming that 'rawVideo' is a video path
cap = cv2.VideoCapture(rawVideo)
#ie, test with rawVideo = 'Data/Easy/TheMartian.mp4'
frame_width = int(cap.get(3))
frame_height = int(cap.get(4))
num_frames = int(cap.get(7))
#create an array that holds all the frames in the video, format: frame_number x width x height
frames = np.zeros([num_frames, frame_height, frame_width, 3], np.uint8)
f = 0
while (cap.isOpened()):
ret, frame = cap.read()
#cv2.imshow('fr ame')
frames[f, :, :, :] = frame
f += 1
if cv2.waitKey(1) & 0xFF == ord('q'):
break
if f == num_frames - 1:
break
cap.release()
cv2.destroyAllWindows()
#need to detect face only on the first frame of the video
#if "good" face not found, then try other frames, until a good face is found
face_found = False
f = 0
face = None
while not face_found:
#face is a Fx4x2 bounding box
face = detectFace(frames[f, :, :, :])
f += 1
if face is not None:
face_found = True
f -= 1
#find the start features
init_frame = f
init_img = frames[init_frame, :, :, :]
init_img_gray = cv2.cvtColor(init_img, cv2.COLOR_BGR2GRAY)
#there will always be 1000 xy's, because we have padded with (0,0)
#make srue to ignore the (0,0) points later in the code
startXs, startYs = getFeatures(init_img_gray, face)
#initialize the the output matrix of tracked images
outputMatrix = np.zeros((num_frames - f, frame_height, frame_width, 3),
np.uint8)
#draw rectangles of all the faces on the current image
initImgWithBBox = init_img
[numFaces, _, _] = face.shape
for i in range(0, numFaces):
bboxOfCurrFace = face[i, :, :]
# get the position of the corners of the bounding box for the current face
first = bboxOfCurrFace[0, :]
second = bboxOfCurrFace[3, :]
# add a bounding box to the initial image
cv2.rectangle(initImgWithBBox, (first[0], first[1]),
(second[0], second[1]), (255, 0, 0))
initImgWithBBox = plotPoints(initImgWithBBox, startYs[:, i],
startXs[:, i])
#add the initial image as the first image
outputMatrix[0, :, :, :] = initImgWithBBox
#actually do the transform and find the new bounding box
for frame in range(f, num_frames - 1): #this should probably not be -1
#get the two consecutive frames at the index
img1 = frames[frame, :, :, :]
img2 = frames[frame + 1, :, :, :]
#find new feature points every 10th frame
if (frame % 10 == 0):
faceCurr = detectFace(img1)
if faceCurr is not None:
face = faceCurr
#else
#just use the last face since it couldn't detect a new face
#convert first image to grey
img1grey = cv2.cvtColor(img1, cv2.COLOR_BGR2GRAY)
#find the starting features on the first image
startXs, startYs = getFeatures(img1grey, face)
[newXs, newYs] = estimateAllTranslation(startXs, startYs, img1, img2)
[Xs, Ys,
newbbox] = applyGeometricTransformation(startXs, startYs, newXs,
newYs, face)
#print newXs
#now add a rectangle of newbbox to img2 and add the feature points
img2WithBoundingBox = img2
for facei in range(0, numFaces):
#get the bounding box for the current face
bboxOfCurrFace = newbbox[facei, :, :]
#get the positions of the two corners for the bounding box of the current face
first = bboxOfCurrFace[0, :].astype(int)
second = bboxOfCurrFace[3, :].astype(int)
#draw the bounding box
img2WithBoundingBox = cv2.rectangle(img2WithBoundingBox,
(first[0], first[1]),
(second[0], second[1]),
(255, 0, 0))
#draw the feature points
if numFaces == 1:
img2WithBoundingBox = plotPoints(img2WithBoundingBox, Ys, Xs)
else:
img2WithBoundingBox = plotPoints(img2WithBoundingBox,
Ys[:, facei], Xs[:, facei])
#add img2 to the output matrix
outputMatrix[frame + 1, :, :, :] = img2WithBoundingBox
#set the new bbox to the face for the next iteration
face = newbbox
#set the xs and ys of the features for the new features
startXs = newXs
startYs = newYs
#output the final video
imageio.mimwrite('finalVideo.avi', outputMatrix, fps=30)
trackedVideo = []
return trackedVideo
# plt.axis('off')
# plt.show()
return Xs, Ys, newbbox
if __name__ == '__main__':
cap = cv2.VideoCapture("./Datasets/Difficult/StrangerThings.mp4")
ret, img1 = cap.read()
ret, img2 = cap.read()
cap.release()
tmpimg1 = img1.copy()
tmpimg2 = img2.copy()
bbox = detectFace(img1)
startXs, startYs = getFeatures(img1, bbox)
newXs, newYs = estimateAllTranslation(startXs, startYs, img1, img2)
Xs, Ys, newbbox = applyGeometricTransformation(startXs, startYs, newXs,
newYs, bbox)
for box in bbox:
cv2.rectangle(tmpimg1, (int(box[0][1]), int(box[0][0])),
(int(box[-1][1]), int(box[-1][0])), (0, 255, 0), 3)
plt.figure()
plt.imshow(tmpimg1)
for j in range(len(startYs)):
plt.plot(startYs[j], startXs[j], 'w+')
plt.axis('off')
plt.show()
for newbox in newbbox:
import pickle
from getFeatures import getFeatures
###accuracy over iterations
##infant state
if __name__ == '__main__':
#filename=raw_input("file name: ")
#_file=open(filename,'r')
#file_reader=csv.reader(_file)
states = [2, 3, 5, 6]
for s in range(0, len(states)):
X, Y, pca, temp_dict = getFeatures(windowSize=1, numStates=states[s])
print states[s], " states"
print pca
X = np.array(X).T
Y = np.array(Y)
dictionary = {v: k for k, v in temp_dict.iteritems()}
labels = []
for i in range(0, len(dictionary)):
labels.append(dictionary[i])
X_train, X_test, Y_train, Y_test = train_test_split(X,
Y,
test_size=0.2,
random_state=1)
"""
def objectTracking(rawVideo):
video = cv2.VideoCapture(rawVideo)
success, prevframe = video.read()
prev_grayframe = cv2.cvtColor(prevframe, cv2.COLOR_BGR2GRAY)
''' Get contour image '''
num_objects = int(input("How many objects to track? : "))
all_bbox_corners = np.zeros((num_objects, 4, 2))
for f in range(num_objects):
contour_image = maskImage(prev_grayframe).astype(np.uint8)
image, bbox_corners = getBoundingBox(prev_grayframe, contour_image)
all_bbox_corners[f, :, :] = bbox_corners
xs, ys = gf.getFeatures(prev_grayframe, all_bbox_corners)
# first_frame = prevframe.copy()
# first_frame[xs, ys, :] = np.array([0, 0, 255])
''' Open up output file '''
w = int(video.get(3))
h = int(video.get(4))
fps = video.get(cv2.CAP_PROP_FPS)
out = cv2.VideoWriter("out.avi", cv2.VideoWriter_fourcc(*'MJPG'), fps,
(w, h))
count = 0 # to count out the iteration
while 1:
prev_grayframe = cv2.cvtColor(prevframe, cv2.COLOR_BGR2GRAY)
success, frame = video.read()
if not success: break
grayframe = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
# Every 6 frames, get new features
if count % 6 == 0:
print("iter", count)
more_xs, more_ys = gf.getFeatures(grayframe, all_bbox_corners)
if more_xs is not None:
xs = np.concatenate((xs, more_xs), axis=0)
ys = np.concatenate((ys, more_ys), axis=0)
#print(xs.shape, ys.shape)
newXs, newYs = eat.estimateAllTranslation(xs, ys, prevframe, frame)
finalXs, finalYs, final_bbox = applyGeometricTransformation(
xs, ys, newXs, newYs, all_bbox_corners)
if finalXs is None:
xs, ys = gf.getFeatures(grayframe, all_bbox_corners)
newXs, newYs = eat.estimateAllTranslation(xs, ys, prevframe, frame)
finalXs, finalYs, final_bbox = applyGeometricTransformation(
xs, ys, newXs, newYs, all_bbox_corners)
output = frame.copy()
for f in range(num_objects):
x = int(np.round(final_bbox[f, 0, 0]))
xw = int(np.round(final_bbox[f, 3, 0]))
y = int(np.round(final_bbox[f, 0, 1]))
yh = int(np.round(final_bbox[f, 3, 1]))
cv2.rectangle(output, (x, y), (xw, yh), (0, 255, 0), 2)
out.write(output)
prevframe = frame.copy()
xs, ys, all_bbox_corners = finalXs, finalYs, final_bbox
count += 1
print("done loop")
video.release()
out.release()
return out
frame1 = generate_output_frame(np.copy(img1), bbox)
frame1 = Image.fromarray(frame1)
frame1.save("easy_frame1.jpg")
# plt.imshow(frame1)
# plt.show()
# For debugging: Show the bounding box we've chosen
# plt.imshow(img1)
# for box in bbox:
# for i in range(3):
# plt.plot(box[i: i+2, 0], box[i: i+2, 1], color="red")
# plt.plot([box[0, 0], box[3, 0]], [box[0, 1], box[3, 1]], color="red")
# plt.show()
# Get the features from inside the bounding box
x, y = getFeatures(rgb2gray(img1), bbox)
# For debugging: Show the bounding box and the features inside
# plt.imshow(img1)
# for box in bbox:
# for i in range(3):
# plt.plot(box[i: i+2, 0], box[i: i+2, 1], color="red")
# plt.plot([box[0, 0], box[3, 0]], [box[0, 1], box[3, 1]], color="red")
# for i in range(x.shape[1]):
# plt.scatter(x[i], y[i][:], color="blue")
# plt.show()
nextframe = np.copy(img2)
warped = np.copy(img1)
newXs = np.copy(x)
newYs = np.copy(y)
def objectTracking(rawVideo):
folder_name = str(datetime.now()).replace('-',
'_').replace(':', '_').replace(
'.', '_').replace(' ', '_')
os.mkdir(folder_name + '_bounding_box')
os.mkdir(folder_name + '_trajectory')
n = 2 # Number of bounding boxes
medium = 0
nof = 5 # number of features
if 'medium' in rawVideo:
medium = 1
nof = 10
n = 1
frameExtractor(folder_name=folder_name, video_path=rawVideo, medium=medium)
noframes = len(os.listdir(folder_name))
filename1 = folder_name + '/0.jpg'
img1 = Image.open(filename1).convert('RGB')
img1 = np.array(img1)
gray1 = rgb2gray(img1).astype('float32')
# to draw own bounding box, comment the next 4 lines and uncomment the 72nd line
if 'easy' in rawVideo:
bbox = bboxgen.get_bbox(img1, n)
else:
bbox = np.array([[[276, 457], [276, 518], [350, 518], [347, 456]]])
# bbox = bboxgen.get_bbox(img1, n)
x_index, y_index = getFeatures(gray1, bbox, nof)
bbox = bbox.astype(float)
for i in range(1, noframes):
filename2 = folder_name + '/' + str(i) + '.jpg'
img2 = Image.open(filename2).convert('RGB')
img2 = np.array(img2)
for b in range(len(bbox)):
x_min = np.min(bbox[b, :, 0])
x_max = np.max(bbox[b, :, 0])
y_min = np.min(bbox[b, :, 1])
y_max = np.max(bbox[b, :, 1])
if len(x_index[x_index[:, b] != -1, b]) <= 3:
if x_min < 0 or y_min < 0 or x_max >= img2.shape[
1] or y_max >= img2.shape[0]:
bbox[b] = 0 * bbox[b] - 1
else:
print('recomputing....')
size = x_index[x_index[:, b] != -1, b].shape[0]
x_old_index = x_index[x_index[:, b] != -1, b]
y_old_index = y_index[y_index[:, b] != -1, b]
gray1 = rgb2gray(img1).astype('float32')
x_index, y_index = getFeatures(gray1, bbox.astype(int),
nof)
x_index[0:size, b] = x_old_index
y_index[0:size, b] = y_old_index
newXs, newYs = estimateAllTranslation(x_index, y_index, img1, img2)
x_index[np.where(newXs == -1)] = -1
y_index[np.where(newXs == -1)] = -1
Xs, Ys, newbbox = applyGeometricTransformation(x_index, y_index, newXs,
newYs, bbox)
xmin = []
ymin = []
xmax = []
ymax = []
for b in range(len(newbbox)):
max_x_y = newbbox[b].max(axis=0)
min_x_y = newbbox[b].min(axis=0)
Xs[Xs[:, b] < min_x_y[0], b] = -1
Xs[Xs[:, b] > max_x_y[0], b] = -1
Ys[Ys[:, b] < min_x_y[1], b] = -1
Ys[Ys[:, b] > max_x_y[1], b] = -1
Ys[Xs[:, b] == -1, b] = -1
Xs[Ys[:, b] == -1, b] = -1
xmin.append(min_x_y[0])
xmax.append(max_x_y[0])
ymin.append(min_x_y[1])
ymax.append(max_x_y[1])
x_index = np.copy(Xs)
y_index = np.copy(Ys)
img1 = np.copy(img2)
bbox = np.copy(newbbox)
plot_box(xmin, ymin, xmax, ymax, img2, i, Xs[Xs != -1], Ys[Ys != -1],
folder_name + '_bounding_box')
#plot_trajectory(Xs[Xs != -1], Ys[Ys != -1], img2, i, folder_name + '_trajectory')
makeMovie(folder_name + '_bounding_box', folder_name + '_bb_output_gif',
noframes)
#Assemble features in train set
#==================================
# This is the section that will take the longest time running since it will actually be extracting the features of
# the malware for each file. This could be a lengthy process.
#==================================
CNN_accuracy = 0
NGRAM_accuracy = 0
combAccPercent = 0
numCrossVals = 10
data_cnn = []
data_nGram = []
base = [0, 0, 0, 0, 0, 0, 0, 0, 0]
labels = []
for fileStub in mySet:
featureListNGRAM = getFeatures.getFeatures(fileStub, "NGRAM", info)
featureList = getFeatures.getFeatures(MALWARE_FILE_PATH + fileStub, "CNN", info)
#check for error
if (len(featureList) == 0 or len(featureListNGRAM) == 0):
#print("FEATURE GRAB ERROR")
#print("Not adding file: " + fileStub + "to the train set")
continue
#Add values to data matrices (one matrix per classifier)
data_nGram.append(numpy.asarray(featureListNGRAM))
data_cnn.append(numpy.asarray(featureList))
#Add classes to class matrix
malwareClass = classDictionary[fileStub]
base2 = base[:]
ret, frame2 = cap.read() # get second frame
ret, frame2 = cap.read() # get second frame
ret, frame2 = cap.read() # get second frame
ret, frame2 = cap.read() # get second frame
ret, frame2 = cap.read() # get second frame
ret, frame2 = cap.read() # get second frame
ret, frame2 = cap.read() # get second frame
ret, frame2 = cap.read() # get second frame
ret, frame2 = cap.read() # get second frame
ret, frame2 = cap.read() # get second frame
frame1_gray = cv2.cvtColor(frame1, cv2.COLOR_RGB2GRAY)
frame2_gray = cv2.cvtColor(frame2, cv2.COLOR_RGB2GRAY)
n_object = 1
bbox = np.array([[[291, 187], [405, 187], [291, 267], [405, 267]]])
startXs, startYs = getFeatures(frame1_gray, bbox)
y = bbox[0][0][1]
x = bbox[0][0][0]
pdb.set_trace()
img_req_1 = frame1_gray[y:bbox[0][3][1], x:bbox[0][3][0]]
corners = cv2.goodFeaturesToTrack(img_req_1, 25, 0.1, 5)
corners = np.int0(corners.squeeze())
trans_x = corners[:, 0].copy() + bbox[0][0][0]
trans_y = corners[:, 1].copy() + bbox[0][0][1]
newXs, newYs = estimateAllTranslation(startXs, startYs, frame1, frame2)
import matplotlib.pyplot as plt
from matplotlib.patches import Rectangle
fig, ax = plt.subplots()
diff = np.subtract(frame1_gray.astype(int), frame2_gray.astype(int))
def basicPredictor(vecnumber):
procvec, diagvec, diagdict = getFeatures.getFeatures()
procvec1 = procvec["Medicare"]
diagvec1 = diagvec["Medicare"]
procvec2 = procvec["Medicaid"]
diagvec2 = diagvec["Medicaid"]
procvec3 = procvec["Uninsured"]
diagvec3 = diagvec["Uninsured"]
keylist = procvec1.keys()
keylist.sort()
veclist = []
diaglist = []
statenamelist = []
i = vecnumber
for key in keylist:
vecs = procvec1[key]
countvec = vecs[i]
veclist.append(countvec)
diaglist.append(diagvec1[key])
try:
veclist.append(procvec2[key][i])
diaglist.append(diagvec2[key])
except:
pass
try:
veclist.append(procvec3[key][i])
diaglist.append(diagvec3[key])
except:
pass
statenamelist.append(key)
veclist = np.asarray(veclist)
diaglist = np.asarray(diaglist)
'''
#encodes only top 5 diagnoses for each input vector, does not give good accuracy!!
diagindex=np.argpartition(diaglist,-5,axis=1)[:,-5:]
hotlist=[]
for i in range(len(diaglist)):
hot=np.zeros((len(diaglist[0]),),dtype=float)
hot[diagindex[i]]=1.0
hotlist.append(hot)
hotlist=np.asarray(hotlist)
hotlist=preprocessing.normalize(hotlist)
'''
#print diagindex
hotlist = preprocessing.normalize(diaglist)
standardscaler = preprocessing.StandardScaler()
vecs = standardscaler.fit_transform(veclist)
# for testing how size of training data affects accuracy
sizevec = [0.11, 0.22, 0.33, 0.44, 0.55, 0.66]
error = []
for i in sizevec:
xtrain, xtest, ytrain, ytest = train_test_split(vecs,
hotlist,
test_size=i,
random_state=17)
knn = neighbors.KNeighborsRegressor()
knn.fit(xtrain, ytrain)
predictedys = knn.score(xtest, ytest)
error.append(predictedys)
print error
'''
return newX, newY
if __name__ == '__main__':
# setup video capture
cap = cv2.VideoCapture("./Datasets/Easy/MarquesBrownlee.mp4")
ret, img1 = cap.read()
ret, img2 = cap.read()
cap.release()
img1 = np.array(img1)
img2 = np.array(img2)
bbox = detectFace(img1)
x, y = getFeatures(img1, bbox)
img1_gray = cv2.cvtColor(img1, cv2.COLOR_BGR2GRAY)
img2_gray = cv2.cvtColor(img2, cv2.COLOR_BGR2GRAY)
Ix, Iy = np.gradient(img1_gray)
Ix = np.array(Ix)
Iy = np.array(Iy)
newXs = []
newYs = []
for i in range(len(x)):
for j in range(len(x[i])):
startX = x[i][j]
startY = y[i][j]
newX, newY = estimateFeatureTranslation(startX, startY, Ix, Iy,
img1_gray, img2_gray)
newXs.append(newX)
def main(video_file, output_filename):
imgs = np.array([])
cap = cv2.VideoCapture(video_file)
ret, img1 = cap.read()
img1 = img1[..., ::-1]
h, w, d = img1.shape
display_img = img1.copy()
display_img = cv2.cvtColor(display_img, cv2.COLOR_BGR2RGB)
cv2.namedWindow("Start Frame")
cv2.setMouseCallback("Start Frame", draw_box)
# Loop until the user is done drawing boxes
while True:
cv2.imshow("Start Frame", display_img)
key = cv2.waitKey(0)
if key == ord('q'):
break
# Destroy the drawing window
cv2.destroyAllWindows()
# Show the result
for i in range(int(len(refPt) / 2)):
cv2.rectangle(display_img, refPt[2 * i], refPt[(2 * i) + 1],
(0, 255, 0), 2)
cv2.imshow("Result", display_img)
cv2.waitKey(0)
cv2.destroyAllWindows()
bbox = []
for i in range(int(len(refPt) / 2)):
# Top Left and bottom right
box_corners = np.array([refPt[2 * i], refPt[(2 * i) + 1]])
start_x, start_y, width, height = cv2.boundingRect(box_corners)
# Create the four coordinates for the box and reshape
box = np.array([[start_x, start_y], [start_x + width, start_y],
[start_x + width, start_y + height],
[start_x, start_y + width]])
bbox.append(box)
# Turn it into a numpy array
bbox = np.array(bbox)
orig_box = np.copy(bbox)
centers = np.zeros((len(bbox), 2))
trajectory_indexer = np.zeros((h, w), dtype=bool)
# Get the features from inside the bounding box
x, y = getFeatures(rgb2gray(img1), bbox)
newXs = np.copy(x)
newYs = np.copy(y)
f = 0
frame = generate_output_frame(np.copy(img1), bbox,
np.copy(trajectory_indexer), np.copy(newXs),
np.copy(newYs))
frame = Image.fromarray(frame)
# Store the processed frames so we can turn it into a video later
all_frames = []
all_frames.append(frame)
a = 0
while ret:
f += 1
a += 1
if not f % 8:
print("Frame: ", f)
a = 1
for i in range(len(bbox)):
# xmin = np.sort(bbox[i, :, 0])[0]
# xmax = np.sort(bbox[i, :, 0])[3]
# ymin = np.sort(bbox[i, :, 1])[0]
# ymax = np.sort(bbox[i, :, 1])[3]
# bbox[i, ...] = np.array([xmin, ymin, xmax, ymin, xmax, ymax, xmin, ymax]).reshape(4,2)
orig_box = np.copy(bbox)
x, y = getFeatures(rgb2gray(img1), bbox)
newXs = np.copy(x)
newYs = np.copy(y)
thresh = .1 + .02 * a
ret, img2 = cap.read()
if not ret:
break
img2 = img2[..., ::-1]
iterations = 1
# Get the new feature locations in the next frame
updatex, updatey, x, y = estimateAllTranslation(
newXs, newYs, np.copy(x), np.copy(y), np.copy(img1), np.copy(img2),
np.copy(bbox))
for k in range(len(bbox)):
centers[k] = np.array(
[np.mean(bbox[k, :, 0]),
np.mean(bbox[k, :, 1])]).astype(int)
# Warp the image for the next iteration
newXs, newYs, bbox, warped = applyGeometricTransformation(
np.copy(x), np.copy(y), updatex, updatey, np.copy(orig_box),
np.copy(img1), np.copy(img2), thresh)
for k in range(len(bbox)):
xcen = int(np.mean(bbox[k, :, 0]))
ycen = int(np.mean(bbox[k, :, 1]))
num = int(
max([abs(xcen - centers[k, 0]),
abs(ycen - centers[k, 1])]))
centerx = np.linspace(centers[k, 0], xcen + 1, num).astype(int)
centery = np.linspace(centers[k, 1], ycen + 1, num).astype(int)
if centerx.size > 0 and centery.size > 0:
trajectory_indexer[centery, centerx] = True
trajectory_indexer[centery + 1, centerx] = True
trajectory_indexer[centery, centerx + 1] = True
trajectory_indexer[centery + 1, centerx + 1] = True
else:
trajectory_indexer[ycen, xcen] = True
trajectory_indexer[ycen + 1, xcen] = True
trajectory_indexer[ycen, xcen + 1] = True
trajectory_indexer[ycen + 1, xcen + 1] = True
frame = generate_output_frame(np.copy(img2), bbox,
np.copy(trajectory_indexer),
np.copy(newXs), np.copy(newYs))
frame = Image.fromarray(frame)
# frame.save("medium_frame%d.jpg" % f)
img1 = np.copy(img2)
all_frames.append(frame)
cap.release()
np_frames = np.array(
[cv2.cvtColor(np.array(f), cv2.COLOR_BGR2RGB) for f in all_frames])
gen_video(np.array(np_frames), "{0}.avi".format(output_filename))
def faceTracking(rawVideo):
#TODO: Your code here
ind = 0
trackedVideo = []
nFrames = len(rawVideo)
# detect faces
bbox = detectFace(rawVideo[0])
# detect features
gray_im = rgb2gray(rawVideo[0])
x, y = getFeatures(gray_im, bbox)
for f in range(0, x.shape[1]):
x[:, f] = x[:, f] + bbox[f][0][1] # row coords
y[:, f] = y[:, f] + bbox[f][0][0] # col coords
for i in range(0, nFrames - 1):
if x.shape[0] < 15:
# detect faces
bbox = detectFace(rawVideo[0])
# detect features
gray_im = rgb2gray(rawVideo[0])
x, y = getFeatures(gray_im, bbox)
for f in range(0, x.shape[1]):
x[:, f] = x[:, f] + bbox[f][0][1] # row coords
y[:, f] = y[:, f] + bbox[f][0][0] # col coords
# get frames
img1 = rawVideo[ind]
img2 = rawVideo[ind + 1]
# track features from first frame to second using KLT procedure
newX, newY = estimateAllTranslation(x, y, img1, img2)
# apply resulting transformation
newXs, newYs, bbox = applyGeometricTransformation(
x, y, newX, newY, bbox)
# apply tracked features and bounding box to frames, update output array
for f in range(0, newXs.shape[1]):
im = cv2.rectangle(img2, (int(bbox[f][0][0]), int(bbox[f][0][1])),
(int(bbox[f][2][0]), int(bbox[f][2][1])),
(0, 0, 0), 3)
for j in range(0, len(newXs)):
im = cv2.circle(im, (int(newYs[j][f]), int(newXs[j][f])), 1,
(0, 0, 255), 2)
trackedVideo.append(im)
ind = ind + 1
x = newXs
y = newYs
trackedVideo = np.array(trackedVideo)
return trackedVideo
from estimateAllTranslation import estimateAllTranslation
from applyGeometricTransformation import applyGeometricTransformation
from helper import rgb2gray
import numpy as np
import cv2
ind = 0
trackedVideo = []
nFrames = len(rawVideo)
# detect faces
bbox = detectFace(rawVideo[0])
# detect features
gray_im = rgb2gray(rawVideo[0])
x, y = getFeatures(gray_im, bbox)
for f in range(0, x.shape[1]):
x[:, f] = x[:, f] + bbox[f][0][1] # row coords
y[:, f] = y[:, f] + bbox[f][0][0] # col coords
# detect faces
bbox = detectFace(rawVideo[0])
# detect features
gray_im = rgb2gray(rawVideo[0])
x, y = getFeatures(gray_im, bbox)
for f in range(0, x.shape[1]):
x[:, f] = x[:, f] + bbox[f][0][1] # row coords
y[:, f] = y[:, f] + bbox[f][0][0] # col coords