def connect(self):
"""Connect producer with consumers."""
started = []
for not_started in self.consumers:
started.append(not_started())
inject(self.producer(), broadcast(*started))
def run_test(data, shape=None, dtype=None):
accumulate = Accumulate(shape=shape, dtype=dtype)
inject(data, accumulate())
np.testing.assert_equal(accumulate.items, data)
target_type = np.ndarray if shape else list
self.assertTrue(isinstance(accumulate.items, target_type))
if shape:
self.assertEqual(accumulate.items.dtype, data.dtype)
def __call__(self):
"""Run the acquisition, i.e. connect the producer and consumer."""
LOG.debug("Running acquisition '{}'".format(self))
started = []
for not_started in self.consumers:
started.append(not_started())
inject(self.generator(), broadcast(*started))
def test_frames():
@coroutine
def count():
while True:
yield
count.i += 1
count.i = 0
inject(frames(5, Camera()), count())
assert count.i == 5
def write(self, data=None, dsetname=None):
"""Write a sequence of *data* if specified, otherwise this method turns into a coroutine.
The data set name is given by *dsetname*.
"""
write_coro = self._write_coroutine(dsetname=dsetname)
if data is None:
return write_coro
else:
inject(data, write_coro)
def write(self, data=None, dsetname=None):
"""Write a sequence of *data* if specified, otherwise this method turns into a coroutine.
The data set name is given by *dsetname*.
"""
write_coro = self._write_coroutine(dsetname=dsetname)
if data is None:
return write_coro
else:
inject(data, write_coro)
def plot_exposure_scan(min_exposure=1*q.ms, max_exposure=500*q.ms, num_points=10):
"""
Plot the mean value of the detector image for exposure times between
*min_exposure* and *max_exposure*, use *num_points* data points.
Returns: a tuple with exposure times and corresponding mean values.
"""
accum = Accumulate()
region = Region(camera['exposure_time'], np.linspace(min_exposure, max_exposure, num_points))
with camera.recording():
inject(resolve(get_exposure_result(region)), broadcast(PyplotViewer(style='-o')(), accum()))
return zip(*accum.items)
def compute_rotation_axis(sinogram, initial_step=None, max_iterations=14,
slice_consumer=None, score_consumer=None):
width_2 = sinogram.shape[1] / 2.0
iteration = 0
step = initial_step or width_2 / 2
current = width_2
while step > 1 and iteration < max_iterations:
frm = current - step
to = current + step
div = 2.0 * step / 5.0
axes = (frm, frm + div, current, current + div, to)
scores = []
for axis in axes:
backproject = Backproject(axis)
result = Result()
inject((sinogram, ), backproject(result()))
backproject.wait()
img = result.result
# Other possibilities: sum(abs(img)) or sum(img * heaviside(-img))
score = np.sum(np.abs(np.gradient(img)))
scores.append(score)
if slice_consumer:
slice_consumer.send(img)
if score_consumer:
score_consumer.send(axis * q.px)
current = axes[scores.index(min(scores))]
step /= 2.0
iteration += 1
return current
def test_injection(self):
inject(generator(), self.consume())
self.assertEqual(self.data, 4)
def setUp(self):
self.timer = Timer()
inject([1, 2, 3], self.timer(null()))
def test_process(self):
result = Result()
inject((1,), process(lambda x: -x, result()))
self.assertEqual(-1, result.result)
def center_rotation_axis(camera, motor, initial_motor_step,
num_iterations=2, num_projections=None, flat=None, dark=None):
"""
Center the rotation axis controlled by *motor*.
Use an iterative approach to center the rotation axis. Around *motor*s
current position, we evaluate five points by running a reconstruction.
*rotation_motor* rotates the sample around the tomographic axis.
*num_iterations* controls the final resolution of the step size, halving
each iteration. *flat* is a flat field frame and *dark* is a dark field
frame which will be used for flat correcting the acuired projections.
"""
width_2 = camera.roi_width.magnitude / 2.0
axis_pos = width_2
# Crop the dark and flat
if flat is not None:
middle = flat.shape[0] / 2
flat = flat[middle, :]
if dark is not None:
dark = dark[middle, :]
n = num_projections or tomo_projections_number(camera.roi_width)
angle_step = np.pi / n * q.rad
step = initial_motor_step
current = motor.position
for i in range(num_iterations):
frm = current - step
to = current + step
div = 2.0 * step / 5.0
positions = (frm, frm + div, current, current + div, to)
scores = []
for position in positions:
motor.position = position
backproject = Backproject(axis_pos)
sino_result = Result()
sino_coro = sino_result()
if flat is not None:
sino_coro = flat_correct(flat, sino_coro, dark=dark)
inject(frames(n, camera, callback=lambda: rotation_motor.move(angle_step).join()),
middle_row(sinograms(n, sino_coro)))
sinogram = (sinogram.result[0, :, :], )
result = Result()
m0 = np.mean(np.sum(sinogram[0], axis=1))
inject(sinogram, backproject(result()))
backproject.wait()
img = result.result
# Other possibilities: sum(abs(img)) or sum(img * heaviside(-img))
score = np.sum(np.abs(np.gradient(img))) / m0
scores.append(score)
current = positions[scores.index(min(scores))]
step /= 2.0
def test_coroutine(self):
inject(self.data, self.walker.write(dsetname='foo'))
self.check()
