def extract(self, img_path):
if not os.path.isfile(img_path):
print('File not found:', img_path)
return None
image = pb.load_single_band(img_path, np.uint8)
self.detector.detect(image)
qr_codes = []
for qr in self.detector.detections:
qr_codes.append({
'text': qr.message,
'points': qr.bounds.convert_tuple()
})
return qr_codes
def scan(self, img_path):
detector = pb.FactoryFiducial(np.uint8).qrcode()
#Load image in grayscale
bcv_img = pb.load_single_band(img_path, np.uint8)
detector.detect(bcv_img)
if (len(detector.detections) > 0):
print(detector.detections[0].message)
return len(detector.detections) + len(detector.failures), len(
detector.detections)
def extract(self, img_path):
image = pb.load_single_band(img_path, np.uint8)
self.detector.detect(image)
qr_codes = []
for qr in self.detector.detections:
qr_codes.append({
'type':
'qr',
'text':
qr.message,
'points':
[list(vertex) for vertex in qr.bounds.convert_tuple()]
})
return qr_codes
import pyboof as pb
import numpy as np
data_path = "../../../data/applet/fiducial/image/examples/"
# Load the camera parameters
intrinsic = pb.Intrinsic()
intrinsic.load_xml(data_path+"intrinsic.xml")
config = pb.ConfigFiducialImage()
print "Configuring detector"
detector = pb.FactoryFiducial( np.uint8 ).squareRobust(config,6)
detector.setIntrinsic(intrinsic)
detector.addPattern(pb.load_single_band(data_path+"../patterns/chicken.png",np.uint8),4.0)
detector.addPattern(pb.load_single_band(data_path+"../patterns/yu.png",np.uint8),4.0)
print "Detecting image"
detector.detect(pb.load_single_band(data_path+"image01.jpg",np.uint8))
print "Number Found = "+str(detector.totalFound())
for i in range(detector.totalFound()):
print "=========== Found "+str(i)
fid_to_cam = detector.getFiducialToCamera(i)
# print fid_to_cam
print "Rotation"
print " "+str(fid_to_cam.get_rotation())
print "Translation"
print " "+str(fid_to_cam.get_translation())
start = time.time()
#camera taking the picture
camera = PiCamera()
print(datetime.now()) #this output is for the logfile
#camera.resolution = (2592, 1944)
camera.resolution = (3280, 2464)
#camera.shutter_speed=4000
camera.capture('/home/pi/Desktop/capture.jpg') # saving picture
camera.close()
#pyboof
pb.init_memmap()
gray = pb.load_single_band('/home/pi/Desktop/capture.jpg',
np.uint8) # loading the picture
detector = pb.FactoryFiducial(np.uint8).qrcode() #qrcode detecting
detector.detect(gray)
values = '' #create empty string
response = detector.detections
for qr in response:
values += '{} '.format(qr.message)
print(values) #this output is for the logfile
print(len(response)) #this output is for the logfile
#temperature
#the sensor adresse needs to be matched to the one of the sensor connected
sensor = '/sys/bus/w1/devices/28-01144058eeaa/w1_slave'
data_path = "../data/example/fiducial/image/examples/"
# Load the camera parameters
intrinsic = pb.CameraPinhole()
intrinsic.load(os.path.join(data_path, "intrinsic.yaml"))
configFiducial = pb.ConfigFiducialImage()
configThreshold = pb.ConfigThreshold.create_local(pb.ThresholdType.LOCAL_MEAN,
10)
print("Configuring detector")
detector = pb.FactoryFiducial(np.uint8).square_image(configFiducial,
configThreshold)
detector.set_intrinsic(intrinsic)
detector.add_pattern(pb.load_single_band(
data_path + "../patterns/pentarose.png", np.uint8),
side_length=4.0)
detector.add_pattern(pb.load_single_band(data_path + "../patterns/yu.png",
np.uint8),
side_length=4.0)
print("Detecting image")
detector.detect(
pb.load_single_band(os.path.join(data_path, "image00.jpg"), np.uint8))
print("Number Found = " + str(detector.get_total()))
for i in range(detector.get_total()):
print("=========== Found #" + str(i))
fid_to_cam = detector.get_fiducial_to_camera(i)
print("Pattern ID = " + str(detector.get_id(i)))
#!/usr/bin/env python3
import numpy as np
import pyboof as pb
import time
# This example shows how you can adjust the number of threads that BoofCV will use in the JVM
original = pb.load_single_band('../data/example/outdoors01.jpg', np.uint8)
gaussian = original.createSameShape()
# Let's warm up the JVM.
for i in range(5):
time0 = time.time()
pb.blur_gaussian(original, gaussian, radius=12)
time1 = time.time()
print("Warm up iteration {:.1f} ms".format(1000 * (time1 - time0)))
print()
print()
N = 30
time0 = time.time()
for i in range(N):
pb.blur_gaussian(original, gaussian, radius=12)
time1 = time.time()
print("Time with default threads {:.1f} ms".format(1000 * (time1 - time0)))
pb.set_max_threads(1)
recognizer = pb.FactorySceneRecognition(np.uint8).scene_recognition()
# First we need to create a model so that it knows how to describe a model. BoofCV does provide a
# pre-build model generated from vacation photos. This is fast enough that often its just easier to train it
# on the images you plan to search.
print("Learning the model. This can take a moment or two.")
recognizer.learn_model(list_images)
# Alternatively you can comment out the code above (lines 18 to 24) and load
# a pre-build model by uncommenting the line below
# recognizer = pb.download_default_scene_recognition(np.uint8, "saved_models")
# Now add all the images that we wish to look up
print("Adding images to the database")
for image_file in list_images:
boof_gray = pb.load_single_band(image_file, np.uint8)
recognizer.add_image(image_file, boof_gray)
# Let's look one up and see which images are related
print("Making a query: ", list_images[6])
query_image = pb.load_single_band(list_images[6], np.uint8)
found_matches = recognizer.query(query_image, 5)
# We are expecting 3 matches to be first, then other two will be incorrect/noise
print("len={}".format(len(found_matches)))
print("\nResults:")
for m in found_matches:
print("{:s} error={:f}".format(m["id"], m["error"]))
# Display the results
image_list = [(query_image, "Query")]
import numpy as np
data_path = "../data/example/fiducial/image/examples/"
# Load the camera parameters
intrinsic = pb.Intrinsic()
intrinsic.load_xml(data_path+"intrinsic.xml")
configFiducial = pb.ConfigFiducialImage()
configThreshold = pb.ConfigThreshold.create_local(pb.ThresholdType.LOCAL_SQUARE,10)
print "Configuring detector"
detector = pb.FactoryFiducial( np.uint8 ).squareImage(configFiducial,configThreshold)
detector.setIntrinsic(intrinsic)
detector.addPattern(pb.load_single_band(data_path+"../patterns/pentarose.png",np.uint8),4.0)
detector.addPattern(pb.load_single_band(data_path+"../patterns/yu.png",np.uint8),4.0)
print "Detecting image"
detector.detect(pb.load_single_band(data_path+"image00.jpg",np.uint8))
print "Number Found = "+str(detector.totalFound())
for i in range(detector.totalFound()):
print "=========== Found #"+str(i)
fid_to_cam = detector.getFiducialToCamera(i)
# print fid_to_cam
print "Pattern ID = "+str(detector.get_id(i))
print "Rotation"
print " "+str(fid_to_cam.get_rotation())
print "Translation"
import numpy as np
import pyboof as pb
from pyboof.swing import visualize_matches
# Enable use of memory mapped files for MUCH faster conversion between some python and boofcv data types
pb.init_memmap(5)
# Load two images
image0 = pb.load_single_band("../data/example/stitch/cave_01.jpg", np.uint8)
image1 = pb.load_single_band("../data/example/stitch/cave_02.jpg", np.uint8)
# Set up the SURF fast hessian feature detector. Reduce the number of features it will detect by putting a limit
# on how close two features can be and the maximum number at each scale
config_fh = pb.ConfigFastHessian()
config_fh.extractRadius = 4
config_fh.maxFeaturesPerScale = 300
# Create the detector and use default for everything else
feature_detector = pb.FactoryDetectDescribe(np.uint8).createSurf(config_detect=config_fh)
# Detect features in the first image
locs0, desc0 = feature_detector.detect(image0)
locs1, desc1 = feature_detector.detect(image1)
print "Detected {:4d} features in image 0".format(len(desc0))
print " {:4d} image 1".format(len(desc1))
factory_association = pb.FactoryAssociate()
factory_association.set_score(pb.AssocScoreType.DEFAULT,feature_detector.get_descriptor_type())
associator = factory_association.greedy()
import pyboof as pb
import numpy as np
original = pb.load_single_band('../../../data/applet/outdoors01.jpg', np.uint8)
gaussian = original.createSameShape(
) # useful function which creates a new image of the
mean = original.createSameShape() # same type and shape as the original
# Apply different types of blur to the image
pb.blur_gaussian(original, gaussian, radius=3)
pb.blur_mean(original, mean, radius=3)
# display the results in a single window as a list
image_list = [(original, "original"), (gaussian, "gaussian"), (mean, "mean")]
pb.swing.show_list(image_list, title="Outputs")
import numpy as np
import pyboof as pb
data_path = "../data/example/calibration/stereo/Bumblebee2_Chess/"
# Load the camera parameters
intrinsic = pb.Intrinsic()
intrinsic.load_xml(data_path + "intrinsicLeft.xml")
# Load original image and the undistorted image
original = pb.load_single_band(data_path + "left08.jpg", np.uint8)
undistorted = original.createSameShape()
# Remove distortion and show the results
pb.remove_distortion(original, undistorted, intrinsic)
image_list = [(original, "Original"), (undistorted, "Undistorted")]
pb.swing.show_list(image_list, title="Lens Distortion")
input("Press any key to exit")
#!/usr/bin/env python3
import numpy as np
import pyboof as pb
from pyboof.swing import visualize_matches
# Enable use of memory mapped files for MUCH faster conversion between some python and boofcv data types
pb.init_memmap(5)
# Load two images and camera calibration
image0 = pb.load_single_band("../data/example/stereo/mono_wall_01.jpg",
np.uint8)
image1 = pb.load_single_band("../data/example/stereo/mono_wall_02.jpg",
np.uint8)
intrinsic = pb.CameraPinhole().load(
"../data/example/calibration/mono/Sony_DSC-HX5V_Chess/intrinsic.yaml")
# Set up the SURF fast hessian feature detector. Reduce the number of features it will detect by putting a limit
# on how close two features can be and the maximum number at each scale
config_fh = pb.ConfigFastHessian()
config_fh.extractRadius = 4
config_fh.maxFeaturesPerScale = 300
# Create the detector and use default for everything else
feature_detector = pb.FactoryDetectDescribe(
np.uint8).createSurf(config_detect=config_fh)
# Detect features in the first image
locs0, desc0 = feature_detector.detect(image0)
locs1, desc1 = feature_detector.detect(image1)
import numpy as np
import pyboof as pb
original = pb.load_single_band("../data/example/outdoors01.jpg", np.uint8)
gaussian = original.createSameShape() # useful function which creates a new image of the
mean = original.createSameShape() # same type and shape as the original
# Apply different types of blur to the image
pb.blur_gaussian(original, gaussian, radius=3)
pb.blur_mean(original, mean, radius=3)
# display the results in a single window as a list
image_list = [(original, "original"), (gaussian, "gaussian"), (mean, "mean")]
pb.swing.show_list(image_list, title="Outputs")
raw_input("Press any key to exit")
import numpy as np
data_path = "../data/example/fiducial/image/examples/"
# Load the camera parameters
intrinsic = pb.Intrinsic()
intrinsic.load_xml(data_path+"intrinsic.xml")
configFiducial = pb.ConfigFiducialImage()
configThreshold = pb.ConfigThreshold.create_local(pb.ThresholdType.LOCAL_SQUARE,10)
print("Configuring detector")
detector = pb.FactoryFiducial( np.uint8 ).squareImage(configFiducial,configThreshold)
detector.setIntrinsic(intrinsic)
detector.addPattern(pb.load_single_band(data_path+"../patterns/pentarose.png",np.uint8),4.0)
detector.addPattern(pb.load_single_band(data_path+"../patterns/yu.png",np.uint8),4.0)
print("Detecting image")
detector.detect(pb.load_single_band(data_path+"image00.jpg",np.uint8))
print("Number Found = "+str(detector.totalFound()))
for i in range(detector.totalFound()):
print("=========== Found #"+str(i))
fid_to_cam = detector.getFiducialToCamera(i)
# print fid_to_cam
print("Pattern ID = "+str(detector.get_id(i)))
print("Rotation")
print(" "+str(fid_to_cam.get_rotation()))
print("Translation")
import pyboof as pb
import numpy as np
original = pb.load_single_band('../../../data/applet/outdoors01.jpg',np.uint8)
# Let BoofCV decide on the type of image to store the gradient as
deriv_dtype = pb.gradient_dtype(pb.get_dtype(original))
# Declare the gradient images
derivX = pb.create_single_band(original.getWidth(),original.getHeight(),deriv_dtype)
derivY = pb.create_single_band(original.getWidth(),original.getHeight(),deriv_dtype)
# Compute the results for a few operators and visualize
pb.gradient(original,derivX,derivY,pb.GradientType.SOBEL)
buffered_sobel = pb.swing.colorize_gradient(derivX,derivY)
pb.gradient(original,derivX,derivY,pb.GradientType.PREWITT)
buffered_prewitt = pb.swing.colorize_gradient(derivX,derivY)
pb.gradient(original,derivX,derivY,pb.GradientType.THREE)
buffered_three = pb.swing.colorize_gradient(derivX,derivY)
pb.gradient(original,derivX,derivY,pb.GradientType.TWO0)
buffered_two0 = pb.swing.colorize_gradient(derivX,derivY)
pb.gradient(original,derivX,derivY,pb.GradientType.TWO1)
buffered_two1 = pb.swing.colorize_gradient(derivX,derivY)
# display the results in a single window as a list
image_list = [(original,"original"),
intrinsic = pb.CameraPinhole()
intrinsic.load(os.path.join(data_path, "intrinsic.yaml"))
# Load a pre-built dictionary
config_marker = pb.load_hamming_marker(pb.HammingDictionary.ARUCO_MIP_25h7)
# Tweak the detector. None class be passed in for the detector
config_detector = pb.ConfigFiducialHammingDetector()
config_detector.configThreshold.type = pb.ThresholdType.LOCAL_MEAN
print("Creating the detector")
detector = pb.FactoryFiducial(np.uint8).square_hamming(config_marker,
config_detector)
print("Detecting image")
detector.detect(
pb.load_single_band(os.path.join(data_path, "image01.jpg"), np.uint8))
print("Number Found = " + str(detector.get_total()))
for i in range(detector.get_total()):
print("=========== Found #" + str(i))
fid_to_cam = detector.get_fiducial_to_camera(i)
print("Pattern ID = " + str(detector.get_id(i)))
print("Image Location " + str(detector.get_center(i)))
if detector.is_3d():
print("Rotation")
print(" " + str(fid_to_cam.get_rotation()))
print("Translation")
print(" " + str(fid_to_cam.get_translation()))
#!/usr/bin/env python3
import numpy as np
import pyboof as pb
from pyboof.swing import visualize_lines
# Load an image with strong lines in it
image = pb.load_single_band("../data/example/simple_objects.jpg", np.uint8)
blurred = image.createSameShape()
# Appying a little bit of blur tends to improve the results
pb.blur_gaussian(image, blurred, radius=5)
# There are a few variants of Hough in BoofCV. The variant we will use here uses the image gradient directly
# This is useful you want to find the edges of objects. If you have an image with thin black lines and you want
# to find the lines and not their edges then the binary variant is what you want to use
config_gradient = pb.ConfigHoughGradient(10)
# Detect the lines using several different variants of Hough line detector
results = []
detector = pb.FactoryDetectLine(
np.uint8).houghLinePolar(config_hough=config_gradient)
results.append(("Gradient Polar", detector.detect(blurred)))
detector = pb.FactoryDetectLine(
np.uint8).houghLineFoot(config_hough=config_gradient)
results.append(("Gradient Foot", detector.detect(blurred)))
# Use swing to visualize the results
visualize_lines(image, results)
#!/usr/bin/env python3
import numpy as np
import pyboof as pb
# Detects all the QR Codes in the image and prints their message and location
data_path = "../data/example/fiducial/qrcode/image03.jpg"
detector = pb.FactoryFiducial(np.uint8).qrcode()
image = pb.load_single_band(data_path, np.uint8)
detector.detect(image)
print("Detected a total of {} QR Codes".format(len(detector.detections)))
for qr in detector.detections:
print("Message: " + qr.message)
print(" at: " + str(qr.bounds))
#!/usr/bin/env python3
import numpy as np
import pyboof as pb
original = pb.load_single_band(
'../data/example/fiducial/image/examples/image00.jpg', np.uint8)
binary = pb.create_single_band(original.getWidth(), original.getHeight(),
np.uint8)
algorithms = []
factory = pb.FactoryThresholdBinary(np.uint8)
algorithms.append(("localGaussian", factory.localGaussian(region_width=11)))
algorithms.append(("localSauvola", factory.localSauvola(region_width=11)))
algorithms.append(("localMean", factory.localMean(region_width=11)))
algorithms.append(("localNick", factory.localNick(region_width=11)))
# algorithms.append(("localOtsu" ,factory.localOtsu(region_width=11))) # This can be slow
algorithms.append(("blockMinMax", factory.blockMinMax(region_width=11)))
algorithms.append(("blockMean", factory.blockMean(region_width=11)))
algorithms.append(("blockOtsu", factory.blockOtsu(region_width=11)))
algorithms.append(("globalEntropy", factory.globalEntropy()))
algorithms.append(("globalOtsu", factory.globalOtsu()))
algorithms.append(("globalLi", factory.globalLi()))
algorithms.append(("globalHuang", factory.globalHuang()))
algorithms.append(("globalFixed", factory.globalFixed(threshold=100)))
image_list = [(original, "Original")]
data_path = "../data/example/fiducial/binary/"
# Load the camera parameters
intrinsic = pb.CameraPinhole()
intrinsic.load(os.path.join(data_path, "intrinsic.yaml"))
configFiducial = pb.ConfigFiducialBinary(target_width=0.3)
configThreshold = pb.ConfigThreshold.create_local(pb.ThresholdType.LOCAL_MEAN, 10)
print("Configuring detector")
detector = pb.FactoryFiducial(np.uint8).square_binary(configFiducial, configThreshold)
# Without intrinsics only pattern ID and pixel location can be found
detector.set_intrinsic(intrinsic)
print("Detecting image")
detector.detect(pb.load_single_band(os.path.join(data_path, "image0000.jpg"), np.uint8))
print("Number Found = "+str(detector.get_total()))
for i in range(detector.get_total()):
print("=========== Found #{}".format(i))
fid_to_cam = detector.get_fiducial_to_camera(i)
print("Pattern ID = "+str(detector.get_id(i)))
print("Image Location " + str(detector.get_center(i)))
if detector.is_3d():
print("Rotation")
print(" "+str(fid_to_cam.get_rotation()))
print("Translation")
print(" "+str(fid_to_cam.get_translation()))
import numpy as np
data_path = "../../../data/applet/fiducial/image/examples/"
# Load the camera parameters
intrinsic = pb.Intrinsic()
intrinsic.load_xml(data_path + "intrinsic.xml")
config = pb.ConfigFiducialImage()
print "Configuring detector"
detector = pb.FactoryFiducial(np.uint8).squareRobust(config, 6)
detector.setIntrinsic(intrinsic)
detector.addPattern(
pb.load_single_band(data_path + "../patterns/chicken.png", np.uint8), 4.0)
detector.addPattern(
pb.load_single_band(data_path + "../patterns/yu.png", np.uint8), 4.0)
print "Detecting image"
detector.detect(pb.load_single_band(data_path + "image01.jpg", np.uint8))
print "Number Found = " + str(detector.totalFound())
for i in range(detector.totalFound()):
print "=========== Found " + str(i)
fid_to_cam = detector.getFiducialToCamera(i)
# print fid_to_cam
print "Rotation"
print " " + str(fid_to_cam.get_rotation())
print "Translation"
import numpy as np
import matplotlib.pyplot as plt
import pyboof as pb
# Enable use of memory mapped files for MUCH faster conversion between some python and boofcv data types
pb.init_memmap(5)
# Load two images
image0 = pb.load_single_band("../data/example/stereo/chair01_left.jpg",
np.uint8)
image1 = pb.load_single_band("../data/example/stereo/chair01_right.jpg",
np.uint8)
# Load stereo rectification
stereo_param = pb.StereoParameters()
stereo_param.load(
"../data/example/calibration/stereo/Bumblebee2_Chess/stereo.xml")
# Rectify and undistort the images
model_rectifier = pb.StereoRectification(stereo_param.left, stereo_param.right,
stereo_param.right_to_left)
model_rectifier.all_inside_left()
distort_left = model_rectifier.create_distortion(
pb.ImageType(image0.getImageType()), True)
distort_right = model_rectifier.create_distortion(
pb.ImageType(image0.getImageType()), False)
rect0 = image0.createSameShape()
rect1 = image1.createSameShape()
[0, 20, 0], # y-axis
[20, 20, 0] # (1,1,0) point
], dtype=np.float64)
# Get Camera matrix
mtx, dist = calibrate_camera.get_calibration_results()
# Initialize axis 3D object points
axis = np.float32([[20, 0, 0], # x-axis
[0, 20, 0], # y-axis
[0, 0, 20] # z-axis
]).reshape(-1, 3)
# Get image
detector = pb.FactoryFiducial(np.uint8).qrcode()
image = pb.load_single_band(filename, np.uint8)
# Detect QR codes in image
detector.detect(image)
num_founded_qr = len(detector.detections)
print("Detected a total of {} QR Codes".format(len(detector.detections)))
for qr in detector.detections:
print("Message: "+qr.message)
print(" at: "+str(qr.bounds))
# Construct 2D image points
object_points_1 = detector.detections[0].bounds
object_points_2 = detector.detections[1].bounds
A_imgpts = np.array([
import numpy as np
import pyboof as pb
data_path = "../data/example/calibration/stereo/Bumblebee2_Chess/"
# Load the camera parameters
intrinsic = pb.Intrinsic()
intrinsic.load_xml(data_path+"intrinsicLeft.xml")
# Load original image and the undistorted image
original = pb.load_single_band(data_path+"left08.jpg", np.uint8)
undistorted = original.createSameShape()
# Remove distortion and show the results
pb.remove_distortion(original, undistorted, intrinsic)
image_list = [(original,"Original"), (undistorted,"Undistorted")]
pb.swing.show_list(image_list, title="Lens Distortion")
raw_input("Press any key to exit")
import numpy as np
import pyboof as pb
original = pb.load_single_band('../data/example/fiducial/image/examples/image00.jpg',np.uint8)
binary = pb.create_single_band(original.getWidth(),original.getHeight(),np.uint8)
algorithms = []
factory = pb.FactoryThresholdBinary(np.uint8)
algorithms.append(("localGaussian",factory.localGaussian(radius=5)))
algorithms.append(("localSauvola" ,factory.localSauvola(radius=5)))
algorithms.append(("localSquare" ,factory.localSquare(radius=5)))
algorithms.append(("globalEntropy",factory.globalEntropy()))
algorithms.append(("globalOtsu" ,factory.globalOtsu()))
algorithms.append(("globalFixed" ,factory.globalFixed(threshold=100)))
image_list = [(original, "Original")]
for a in algorithms:
a[1].process(original, binary)
buffered_binary = pb.swing.render_binary(binary)
image_list.append((buffered_binary,a[0]))
pb.swing.show_list(image_list, title="Binary Thresholding")
raw_input("Press any key to exit")
#!/usr/bin/env python3
import numpy as np
import pyboof as pb
from pyboof.swing import visualize_matches
# Enable use of memory mapped files for MUCH faster conversion between some python and boofcv data types
pb.init_memmap()
# Load two images
image0 = pb.load_single_band("../data/example/stitch/cave_01.jpg", np.uint8)
image1 = pb.load_single_band("../data/example/stitch/cave_02.jpg", np.uint8)
# Set up the SURF fast hessian feature detector. Reduce the number of features it will detect by putting a limit
# on how close two features can be and the maximum number at each scale
config_fh = pb.ConfigFastHessian()
config_fh.extractRadius = 4
config_fh.maxFeaturesPerScale = 300
# Create the detector and use default for everything else
feature_detector = pb.FactoryDetectDescribe(np.uint8).createSurf(config_detect=config_fh)
# Detect features in the first image
locs0, desc0 = feature_detector.detect(image0)
locs1, desc1 = feature_detector.detect(image1)
print("Detected {:4d} features in image 0".format(len(desc0)))
print(" {:4d} image 1".format(len(desc1)))
#!/usr/bin/env python3
import numpy as np
import os
import pyboof as pb
data_path = "../data/example/calibration/stereo/Bumblebee2_Chess/"
# Load the camera parameters
intrinsic = pb.CameraPinhole()
intrinsic.load(os.path.join(data_path, "intrinsicLeft.yaml"))
# Load original image and the undistorted image
original = pb.load_single_band(os.path.join(data_path, "left08.jpg"), np.uint8)
undistorted = original.createSameShape()
# Remove distortion and show the results
pb.remove_distortion(original, undistorted, intrinsic)
image_list = [(original, "Original"), (undistorted, "Undistorted")]
pb.swing.show_list(image_list, title="Lens Distortion")
input("Press any key to exit")
import numpy as np
import cv2
import pyboof as pb
# Enable use of memory mapped files for MUCH faster conversion of images between java and python
pb.init_memmap(5)
image_path = '../data/example/outdoors01.jpg'
# Can load an image using OpenCV then convert it into BoofCV
ndarray_img = cv2.imread(image_path, 0)
boof_cv = pb.ndarray_to_boof(ndarray_img)
# Can also use BoofCV to load the image directly
boof_gray = pb.load_single_band(image_path, np.uint8)
boof_color = pb.load_planar(image_path, np.uint8)
# Let's display all 3 of them in Java
# display the results in a single window as a list
image_list = [(boof_cv, "OpenCV"), (boof_gray, "Gray Scale"),
(boof_color, "Color")]
pb.swing.show_list(image_list, title="Images")
input("Press any key to exit")
import numpy as np
import cv2
import pyboof as pb
# Enable use of memory mapped files for MUCH faster conversion of images between java and python
pb.init_memmap(5)
image_path = '../data/example/outdoors01.jpg'
# Can load an image using OpenCV then convert it into BoofCV
ndarray_img = cv2.imread(image_path,0)
boof_cv = pb.ndarray_to_boof(ndarray_img)
# Can also use BoofCV to load the image directly
boof_gray = pb.load_single_band(image_path,np.uint8)
boof_color = pb.load_planar(image_path,np.uint8)
# Let's display all 3 of them in Java
# display the results in a single window as a list
image_list = [(boof_cv,"OpenCV"),
(boof_gray,"Gray Scale"),
(boof_color,"Color")]
pb.swing.show_list(image_list,title="Images")
raw_input("Press any key to exit")
import pyboof as pb
import numpy as np
import os
import glob
# Demonstration of how to calibrate a camera using a pinhole model
data_path = "../data/example/calibration/stereo/Bumblebee2_Chess/"
print("Configuring and creating a chessboard detector")
config_grid = pb.ConfigGridDimen(num_rows=5, num_cols=7, square_width=0.3)
detector = pb.FactoryFiducialCalibration.chessboardX(config_grid)
print("Detecting calibration targets")
observations = []
for file in glob.glob(os.path.join(data_path, "left*.jpg")):
image = pb.load_single_band(file, np.float32)
detector.detect(image)
if detector.detected_points:
print("success " + file)
o = {
"width": image.getWidth(),
"height": image.getHeight(),
"pixels": detector.detected_points
}
observations.append(o)
else:
print("failed " + file)
print("Solving for intrinsic parameters")
intrinsic, errors = pb.calibrate_brown(observations,
# Demonstration of how to calibrate a stereo camera
data_path = "../data/example/calibration/stereo/Zed_ecocheck/"
print("Configuring and creating a chessboard detector")
config_ecocheck = pb.ecocheck_parse("9x7n1", square_size=0.3)
detector = pb.FactoryFiducialCalibration.ecocheck(config_ecocheck)
print("Detecting calibration targets")
files_left = sorted(glob.glob(os.path.join(data_path, "left*.jpg")))
files_right = sorted(glob.glob(os.path.join(data_path, "right*.jpg")))
observations_left = []
observations_right = []
for file_left, file_right in zip(files_left, files_right):
# left image
image = pb.load_single_band(file_left, np.float32)
detector.detect(image)
o = {"width": image.getWidth(), "height": image.getHeight()}
if detector.detected_markers:
print("success " + file_left)
o["pixels"] = detector.detected_markers[0]["landmarks"]
else:
o["pixels"] = []
observations_left.append(o)
# right image
image = pb.load_single_band(file_right, np.float32)
detector.detect(image)
o = {"width": image.getWidth(), "height": image.getHeight()}
if detector.detected_markers:
print("success " + file_right)
#!/usr/bin/env python3
import pyboof as pb
import numpy as np
import os
import cv2
# Demonstration of how to detect points in a calibration target
data_path = "../data/example/calibration/stereo/Bumblebee2_Chess/"
print("Configuring and creating a chessboard detector")
config_grid = pb.ConfigGridDimen(num_rows=5, num_cols=7, square_width=0.3)
detector = pb.FactoryFiducialCalibration.chessboardB(config_grid)
print("Detecting image")
image = pb.load_single_band(os.path.join(data_path, "left01.jpg"), np.float32)
detector.detect(image)
print("Detected points {}".format(len(detector.detected_points)))
# Convert it into a color image for visualization purposes
ndimage = pb.boof_to_ndarray(image).astype(np.uint8)
ndimage = cv2.cvtColor(ndimage, cv2.COLOR_GRAY2RGB)
# Draw green dots with red outlines
for x in detector.detected_points:
# x[0] is index of the control point
# x[1] and x[2] is the (x,y) pixel coordinate, sub-pixel precision
cv2.circle(ndimage, (int(x[1]), int(x[2])), 7, (0, 0, 255), -1)
cv2.circle(ndimage, (int(x[1]), int(x[2])), 5, (0, 255, 0), -1)
import numpy as np
import matplotlib.pyplot as plt
import pyboof as pb
# Enable use of memory mapped files for MUCH faster conversion between some python and boofcv data types
pb.init_memmap(5)
# Load two images
image0 = pb.load_single_band("../data/example/stereo/chair01_left.jpg", np.uint8)
image1 = pb.load_single_band("../data/example/stereo/chair01_right.jpg", np.uint8)
# Load stereo rectification
stereo_param = pb.StereoParameters()
stereo_param.load("../data/example/calibration/stereo/Bumblebee2_Chess/stereo.xml")
# Rectify and undistort the images
model_rectifier = pb.StereoRectification(stereo_param.left, stereo_param.right, stereo_param.right_to_left)
model_rectifier.all_inside_left()
distort_left = model_rectifier.create_distortion(pb.ImageType(image0.getImageType()), True)
distort_right = model_rectifier.create_distortion(pb.ImageType(image0.getImageType()), False)
rect0 = image0.createSameShape()
rect1 = image1.createSameShape()
distort_left.apply(image0, rect0)
distort_right.apply(image1, rect1)
# Configure and compute disparity
config = pb.ConfigStereoDisparity()