def main():
(image1_path, image2_path) = optcheck.getArguments()
image1 = highgui.openImage(image1_path)
image2 = highgui.openImage(image2_path)
imgproc.detectCorners(image1)
imgproc.detectCorners(image2)
highgui.showImagesHorizontally("Corner feature detection", image1, image2)
highgui.saveImage(image1, highgui.getSavePath(image1_path, '12'))
highgui.saveImage(image2, highgui.getSavePath(image2_path, '12'))
def main():
imagePath = optcheck.getArguments()[0]
image = highgui.openImage(imagePath)
image = imgproc.detectLines(image, 100, 200)
highgui.showImage("name", image)
highgui.saveImage(image, highgui.getSavePath(imagePath, '11'))
def onMouse(event, x, y, flags, userdata):
global mouseClicks, image, windowName
print("sure buddy")
if (cv2.EVENT_LBUTTONDOWN == event and mouseClicks < 4):
src[mouseClicks] = [x, y]
mouseClicks += 1
if (mouseClicks == 4):
mouseClicks += 1 # increase one more time so this function will basically do nothing anymore
M = cv2.getPerspectiveTransform(src,
dst) # Get the transformation matrix
warped = cv2.warpPerspective(image, M, (-1, -1)) # warp the image
result = cv2.hconcat(
(image, warped)) # concatenate the original image and warped image
cv2.imshow(windowName, result) # show original and warped image
# Saving the warped image
highgui.saveImage(warped, highgui.getSavePath(imagepath, '9'))
def onMouse(event, x, y, flags, userdata):
global mouseClicks, image, windowName, imagePath
if (cv2.EVENT_LBUTTONDOWN == event and mouseClicks < 4):
src[mouseClicks] = [x, y]
mouseClicks += 1
if (mouseClicks == 4):
mouseClicks += 1 # increase one more time so this function will basically do nothing anymore
for i in range(0, 3):
cv2.line(img=image,
pt1=(src[i][0], src[i][1]),
pt2=(src[i + 1][0], src[i + 1][1]),
color=(0, 0, 255),
thickness=10)
cv2.line(img=image,
pt1=(src[3][0], src[3][1]),
pt2=(src[0][0], src[0][1]),
color=(0, 0, 255),
thickness=10)
print(filename, end='')
for point in src:
print(f" ({point[0]}, {point[1]}) ", end='')
print("\n")
cv2.imwrite(highgui.getSavePath(imagePath, 'e'), image)
out = [imagePath, src[0], src[1], src[2], src[3]]
config = Path('GT.txt')
if config.is_file() == False:
config.touch()
pickled = []
pickled.append(out)
with open('GT.txt', 'wb') as f:
pickle.dump(pickled, f)
else:
pickled = []
with open('GT.txt', 'rb') as f:
pickled = pickle.load(f)
pickled.append(out)
with open('GT.txt', 'wb') as f:
pickle.dump(pickled, f)
def evaluateClassifierPerformance(classifier):
avg_precision = 0
avg_recall = 0
# apply 4-fold cross validation: image with index i will be used as testdata, the other images as trainingdata
for i in range(1, len(IMAGES) + 1):
trainingSet = IMAGES[:i - 1] + IMAGES[i:]
testImage = IMAGES[i - 1]
featureSet = []
labelSet = []
for imagePath in trainingSet:
image = highgui.openImage(f"{imagePath}.png")
image_blocks = highgui.openImage(f"{imagePath}_blocks.png")
features = getFeatureVectors(image)
labels = getRoadMarkings(image_blocks)
(features, labels) = balanceTrainingset(features, labels)
featureSet.extend(features)
labelSet.extend(labels)
classifier.fit(featureSet, labelSet)
image = highgui.openImage(f"{testImage}.png")
image_blocks = highgui.openImage(f"{testImage}_blocks.png")
predictions = classifier.predict(getFeatureVectors(image))
groundTruth = getRoadMarkings(image_blocks)
# visualize the filter means
visualizeMeans(getFeatureVectors(image), groundTruth,
f"{IMAGES[i - 1]}_filter.png")
predictionImage = getOverlayedImage(image, predictions)
highgui.saveImage(
predictionImage,
highgui.getSavePath(f"{testImage}.png", 'CLASSIFIER'))
TP, FP, FN, TN = (0, 0, 0, 0)
for prediction, groundTruth in zip(predictions, groundTruth):
if (groundTruth == 1):
if (prediction == 1):
TP += 1
else:
FN += 1
elif (groundTruth == 0):
if (prediction == 1):
FP += 1
else:
TN += 1
precision = round(
TP / (TP + FP), 4
) * 100 # given a positive prediction from the classifier, how likely is it to be correct?
recall = round(
TP / (TP + FN), 4
) * 100 # given a positive example, will the classifier detect it?
avg_precision += precision
avg_recall += recall
#print(f"road{i} & {precision}\\% & {recall}\\% \\\\") # for latex
print(
f"recall and precision considering {trainingSet} as the training data and {testImage} as the test data"
)
print(f"Precision: {precision} \t Recall: {recall}")
def main():
imagePath = optcheck.getArguments()[0]
image = highgui.openImage(imagePath)
result = imgproc.extractEdges(image, -15)
highgui.showImage(imagePath, result)
highgui.saveImage(result, highgui.getSavePath(imagePath, '10'))
import cv2
from Modules import optcheck, highgui, imgproc
#rainbowdiscs.png
imagePath = optcheck.getArguments()[0]
image = highgui.openImage(imagePath)
D = cv2.getStructuringElement(cv2.MORPH_RECT, (1, 5))
dilation = cv2.dilate(image, D, iterations=3)
E = cv2.getStructuringElement(cv2.MORPH_RECT, (3, 3))
erosion = cv2.erode(dilation, E, iterations=3)
highgui.showImagesHorizontally("EX7", image, dilation, erosion)
highgui.saveImage(erosion, highgui.getSavePath(imagePath, '7EROSION'))
highgui.saveImage(dilation, highgui.getSavePath(imagePath, '7DILATION'))
import cv2
import numpy
from Modules import optcheck, highgui, imgproc
#shadow.png
imagepath = optcheck.getArguments()[0]
image = highgui.openImage(imagepath)
shear_factor = -0.2
M = numpy.float32([[1, shear_factor, 50], [0, 1, 0]])
shearImage = cv2.warpAffine(image, M, (-1, -1))
highgui.showImagesHorizontally("Shearing", image, shearImage)
highgui.saveImage(shearImage, highgui.getSavePath(imagepath, '8'))