def test_afLC_ds_u16_nm():
downsample = 4
afc = afLC.AFLockC(offset = 0.0, downsample = downsample)
cx = 32.0
cy = 64.0
for i in range(10):
x1_off = cx + 10.0 * (random.random() - 0.5)
y1_off = cy + 40.0 * (random.random() - 0.5)
x2_off = 3*cx + 10.0 * (random.random() - 0.5)
y2_off = cy + 40.0 * (random.random() - 0.5)
im = dg.drawGaussiansXY((128,128),
numpy.array([x1_off, x2_off]),
numpy.array([y1_off, y2_off]),
sigma = downsample)
im = (100.0*im).astype(numpy.uint16)
[dx, dy, res, mag] = afc.findOffsetU16NM(im)
assert res, "Fitting failed."
assert(numpy.allclose(numpy.array([downsample*dx, downsample*dy]),
numpy.array([x1_off - x2_off + 2.0*cx, y1_off - y2_off]),
atol = 1.0e-2,
rtol = 1.0e-2))
def __init__(self, parameters=None, **kwds):
kwds["parameters"] = parameters
super().__init__(**kwds)
self.cnt = 0
self.max_backlog = 20
self.min_good = parameters.get("min_good")
self.reps = parameters.get("reps")
self.sum_scale = parameters.get("sum_scale")
self.sum_zero = parameters.get("sum_zero")
self.good = numpy.zeros(self.reps, dtype=numpy.bool)
self.mag = numpy.zeros(self.reps)
self.x_off = numpy.zeros(self.reps)
self.y_off = numpy.zeros(self.reps)
# Create slices for selecting the appropriate regions from the camera.
t1 = list(map(int, parameters.get("roi1").split(",")))
self.roi1 = (slice(t1[0], t1[1]), slice(t1[2], t1[3]))
t2 = list(map(int, parameters.get("roi2").split(",")))
self.roi2 = (slice(t2[0], t2[1]), slice(t2[2], t2[3]))
self.afc = afLC.AFLockC(offset=parameters.get("background"),
downsample=parameters.get("downsample"))
assert (self.reps >= self.min_good), "'reps' must be >= 'min_good'."
def test_afLC():
afc = afLC.AFLockC(offset=0.0)
cx = 16.0
cy = 32.0
for i in range(10):
x1_off = cx + 10.0 * (random.random() - 0.5)
y1_off = cy + 40.0 * (random.random() - 0.5)
x2_off = cx + 10.0 * (random.random() - 0.5)
y2_off = cy + 40.0 * (random.random() - 0.5)
im1 = dg.drawGaussiansXY((32, 64), numpy.array([x1_off]),
numpy.array([y1_off]))
im2 = dg.drawGaussiansXY((32, 64), numpy.array([x2_off]),
numpy.array([y2_off]))
[dx, dy, res, mag] = afc.findOffset(im1, im2)
assert (res.success)
assert (numpy.allclose(numpy.array([dx, dy]),
numpy.array([x1_off - x2_off, y1_off - y2_off]),
atol=1.0e-3,
rtol=1.0e-3))
def test_afLC_ds():
downsample = 4
afc = afLC.AFLockC(offset=0.0, downsample=downsample)
cx = 32.0
cy = 64.0
for i in range(10):
x1_off = cx + 10.0 * (random.random() - 0.5)
y1_off = cy + 40.0 * (random.random() - 0.5)
x2_off = cx + 10.0 * (random.random() - 0.5)
y2_off = cy + 40.0 * (random.random() - 0.5)
im1 = dg.drawGaussiansXY((64, 128),
numpy.array([x1_off]),
numpy.array([y1_off]),
sigma=downsample)
im2 = dg.drawGaussiansXY((64, 128),
numpy.array([x2_off]),
numpy.array([y2_off]),
sigma=downsample)
[dx, dy, res, mag] = afc.findOffset(im1, im2)
assert (res.success)
assert (numpy.allclose(numpy.array([downsample * dx, downsample * dy]),
numpy.array([x1_off - x2_off, y1_off - y2_off]),
atol=1.0e-2,
rtol=1.0e-2))
def test_afLC_offset():
afc = afLC.AFLockC(offset = 0.0)
cx = 16.0
cy = 32.0
for i in range(10):
x1_off = cx + random.randint(-5, 5)
y1_off = cy + random.randint(-20, 20)
x2_off = cx + random.randint(-5, 5)
y2_off = cy + random.randint(-20, 20)
im1 = dg.drawGaussiansXY((32,64), numpy.array([x1_off]), numpy.array([y1_off]))
im2 = dg.drawGaussiansXY((32,64), numpy.array([x2_off]), numpy.array([y2_off]))
afc.findOffset(im1, im2)
offset = afc.getOffset()
assert(numpy.allclose(offset, numpy.array([x1_off - x2_off, y1_off - y2_off])))
def __init__(self, parameters=None, **kwds):
kwds["parameters"] = parameters
super().__init__(**kwds)
self.cnt = 0
self.max_backlog = 20
self.min_good = parameters.get("min_good")
self.reps = parameters.get("reps")
self.sum_scale = parameters.get("sum_scale")
self.sum_zero = parameters.get("sum_zero")
self.bg_est = numpy.zeros(self.reps)
self.good = numpy.zeros(self.reps, dtype=numpy.bool)
self.mag = numpy.zeros(self.reps)
self.x_off = numpy.zeros(self.reps)
self.y_off = numpy.zeros(self.reps)
self.afc = afLC.AFLockC(offset=parameters.get("background"),
downsample=parameters.get("downsample"))
assert (self.reps >= self.min_good), "'reps' must be >= 'min_good'."
def test_hess():
dx = 1.0e-6
afc_py = afLC.AFLockPy(offset=0.0)
afc = afLC.AFLockC(offset=0.0)
x1_off = 8.0
y1_off = 16.0
x2_off = 8.0
y2_off = 16.0
im1 = dg.drawGaussiansXY((32, 64), numpy.array([x1_off]),
numpy.array([y1_off]))
im2 = dg.drawGaussiansXY((32, 64), numpy.array([x2_off]),
numpy.array([y2_off]))
# Initialize fitter.
afc_py.findOffset(im1, im2)
afc.findOffset(im1, im2)
for i in range(10):
v1 = numpy.random.normal(size=2)
# Exact.
hce_py = afc_py.hessCost(v1)
hce_c = afc.hessCost(v1)
# Analytic.
hca = numpy.zeros((2, 2))
v2 = numpy.copy(v1)
v2[0] += dx
hca[0, :] = (afc_py.gradCost(v2) - afc_py.gradCost(v1)) / dx
v2 = numpy.copy(v1)
v2[1] += dx
hca[1, :] = (afc_py.gradCost(v2) - afc_py.gradCost(v1)) / dx
assert (numpy.allclose(hca, hce_py, atol=1.0e-3, rtol=1.0e-3))
assert (numpy.allclose(hca, hce_c, atol=1.0e-3, rtol=1.0e-3))