def configure(self, p, i, o):
self.points = MatReader(filename='%s/points.yaml' % p.directory)
self.points3d = MatReader(filename='%s/points3d.yaml' % p.directory)
self.descriptors = MatReader(filename='%s/descriptors.yaml' % p.directory)
self.R = MatReader(filename='%s/R.yaml' % p.directory)
self.T = MatReader(filename='%s/T.yaml' % p.directory)
self.image = imread(image_file='%s/train.png' % p.directory)
def configure(self, p, i, o):
self.points = MatReader(filename='%s/points.yaml' % p.directory)
self.points3d = MatReader(filename='%s/points3d.yaml' % p.directory)
self.descriptors = MatReader(filename='%s/descriptors.yaml' % p.directory)
self.R = MatReader(filename='%s/R.yaml' % p.directory)
self.T = MatReader(filename='%s/T.yaml' % p.directory)
self.image = imread(image_file='%s/train.png' % p.directory)
def match(options,
scale=1.0,
train_scale=1 / 3.0,
scale_factor=1.05,
n_levels=5,
n_features_train=10000,
n_features_test=1000,
match_distance=120,
):
if not os.path.exists(options.output):
os.makedirs(options.output)
train = imread('Train image', image_file=options.train,
mode=ImageMode.GRAYSCALE)
test = imread('Test image', image_file=options.test,
mode=ImageMode.GRAYSCALE)
train_name, _extension = os.path.splitext(os.path.basename(options.train))
test_name, _extension = os.path.splitext(os.path.basename(options.test))
print 'Scale %d is %f' % (0, scale)
for i in range(n_levels - 1):
scale *= 1.0 / scale_factor
print 'Scale %d is %f' % (i + 1, scale)
orb_train = ORB(n_features=n_features_train,
n_levels=n_levels, scale_factor=scale_factor)
orb_test = ORB(n_features=n_features_test, n_levels=1)
plasm = ecto.Plasm()
scale = Scale(factor=train_scale, interpolation=AREA)
plasm.connect(train['image'] >> scale['image'])
train_pts = hookup_orb(plasm, scale, orb_train)
test_pts = hookup_orb(plasm, test, orb_test)
matcher = Matcher()
plasm.connect(orb_train['descriptors'] >> matcher['train'],
orb_test['descriptors'] >> matcher['test'],
)
matchesMat = MatchesToMat()
h_est = MatchRefinement(match_distance=match_distance)
plasm.connect(train_pts['points'] >> h_est['train'],
test_pts['points'] >> h_est['test'],
matcher['matches'] >> h_est['matches'],
h_est['matches'] >> matchesMat['matches']
)
match_draw = DrawMatches()
output_base = os.path.join(options.output,
'%(train)s_%(test)s' %
dict(train=train_name, test=test_name))
match_image = output_base + '.png'
image_saver = ImageSaver(filename=match_image)
plasm.connect(train_pts['points'] >> match_draw['train'],
test_pts['points'] >> match_draw['test'],
h_est['matches'] >> match_draw['matches'],
h_est['matches_mask'] >> match_draw['matches_mask'],
scale['image'] >> match_draw['train_image'],
test['image'] >> match_draw['test_image'],
match_draw[:] >> image_saver[:]
)
match_eval = MatchEval(output=options.output)
plasm.connect(matchesMat['matches'] >> match_eval['matches'],
train_pts['points'] >> match_eval['train'],
test_pts['points'] >> match_eval['test'],
h_est['matches_mask'] >> match_eval['inliers'],
)
run_plasm(options, plasm, locals=vars())
def do_projector():
options = parse_options()
# define the input
subs = dict(image=ImageSub(topic_name='camera/rgb/image_color', queue_size=1),
depth=ImageSub(topic_name='camera/depth/image', queue_size=1),
#depth_info=CameraInfoSub(topic_name='camera/depth/camera_info', queue_size=0),
#image_info=CameraInfoSub(topic_name='camera/rgb/camera_info', queue_size=0),
)
sync = ecto_ros.Synchronizer('Synchronizator', subs=subs)
im2mat_rgb = ecto_ros.Image2Mat()
im2mat_depth = ecto_ros.Image2Mat()
invert = imgproc.BitwiseNot()
brg2rgb = imgproc.cvtColor('bgr -> rgb', flag=4)
bgr2gray = imgproc.cvtColor('bgr -> gray', flag=7)
camera_info = calib.CameraIntrinsics('Camera Info', camera_file="camera.yml")
calibrator = projector.Calibrator()
plasm = ecto.Plasm()
offset_x = 0
s1 = ecto.Strand()
main_display = highgui.imshow("rgb show", name="rgb", waitKey=5, strand=s1)
graph = [
sync["image"] >> im2mat_rgb["image"],
im2mat_rgb["image"] >> main_display[:],
sync["depth"] >> im2mat_depth["image"],
im2mat_depth["image"] >> highgui.imshow("depth show", name="depth", waitKey= -1, strand=s1)[:],
im2mat_rgb["image"] >> (brg2rgb["input"], bgr2gray["input"]),
# sync["image","depth"] >> pattern_detection['image', 'depth'],
# pattern_detection['points'] >> projection_estimator['points']
]
print "Press 's' to capture a view."
# add the display of the pattern
video_display = highgui.imshow('pattern',
name='video_cap', waitKey=2, maximize=False, autoSize=True)
case = 2
if case == 0:
offset_x = -.25
pose_from_fiducial = PoseFromFiducial(plasm,
rows=5, cols=3,
pattern_type="acircles",
square_size=0.04,
offset_x=offset_x,
debug=DEBUG)
# deal with fiducial markers
graph += [ brg2rgb["out"] >> pose_from_fiducial['color_image'],
bgr2gray["out"] >> pose_from_fiducial['image'],
camera_info['K'] >> pose_from_fiducial['K'],
pose_from_fiducial['debug_image'] >> highgui.imshow("pattern show", name="pattern", waitKey= -1, strand=s1)[:],
]
# Deal with the warping
warper = projector.FiducialWarper(projection_file='projector_calibration.yml',
radius=0.10)
graph += [pose_from_fiducial['R', 'T', 'found'] >> warper['R', 'T', 'found'],
warper['output'] >> highgui.imshow("warped image", name="warped", waitKey= -1, strand=s1)[:],
]
elif case == 1:
warper = projector.DepthWarper(projection_file='projector_calibration.yml')
graph += [camera_info['K'] >> warper['K'],
im2mat_depth["image"] >> warper['depth'],
warper['output'] >> highgui.imshow("warped image", name="warped", waitKey= -1, strand=s1)[:],
]
elif case == 2:
# Deal with the warping
warper = projector.ImageWarper(projection_file='projector_calibration.yml', offset_x=0,offset_y=0)
pose_from_plane = projector.PlaneFitter()
pose_draw = calib.PoseDrawer('Plane Pose Draw')
graph += [im2mat_depth["image"] >> pose_from_plane['depth'],
camera_info['K'] >> pose_from_plane['K'],
pose_from_plane['R', 'T'] >> warper['R', 'T'],
im2mat_rgb["image"] >> (pose_draw['image'],),
highgui.imread(image_file='lena.jpg')["image"] >> (warper['image'],),
camera_info['K'] >> pose_draw['K'],
pose_from_plane['R', 'T'] >> pose_draw['R', 'T'],
pose_draw['output'] >> highgui.imshow("pose", name="pose", waitKey= -1, strand=s1)[:],
warper['output'] >> highgui.imshow("warped image", name="warped", waitKey= -1, strand=s1)[:],
]
plasm.connect(graph)
# display DEBUG data if needed
#if DEBUG:
# print plasm.viz()
# ecto.view_plasm(plasm)
# execute the pipeline
sched = ecto.schedulers.Singlethreaded(plasm)
sched.execute()
import argparse
parser = argparse.ArgumentParser(description='Test orb on images.')
parser.add_argument('-i,--input', dest='input',
help='The input dir. %(default)s', default='./images')
parser.add_argument('-o,--output', dest='output', type=str,
help='The output directory for this template. Default: %(default)s', default='./')
factory = cell_options(parser, ORB, 'ORB')
scheduler_options(parser.add_argument_group('Scheduler'), default_niter=1)
options = parser.parse_args()
options.niter = 1
options.orb_factory = factory
return options
options = parse_args()
image = imread(image_file=options.input)
shutil.copy(options.input, os.path.join(options.output, 'train.png'))
orb_m = options.orb_factory(options)
rgb2gray = cvtColor (flag=Conversion.RGB2GRAY)
kpts2mat = KeypointsToMat()
ptsTo3d = PointsTo3d(scale=0.0254 / 100) #100 dpi ~> 25.4 mm/ 100 px
plasm = ecto.Plasm()
plasm.connect(image['image'] >> orb_m['image'],
orb_m['keypoints'] >> kpts2mat['keypoints'],
kpts2mat['points'] >> ptsTo3d['points']
)
if not os.path.exists(options.output):
print 'making ', options.output
os.makedirs(options.output)
