def test_ScoreboardModel_predict_percept():
model = ScoreboardModel(xystep=0.55, rho=100, thresh_percept=0)
model.build()
# Single-electrode stim:
img_stim = np.zeros(60)
img_stim[47] = 1
percept = model.predict_percept(ArgusII(stim=img_stim))
# Single bright pixel, very small Gaussian kernel:
npt.assert_equal(np.sum(percept.data > 0.9), 1)
npt.assert_equal(np.sum(percept.data > 0.5), 1)
npt.assert_equal(np.sum(percept.data > 0.1), 7)
npt.assert_equal(np.sum(percept.data > 0.00001), 35)
# Brightest pixel is in lower right:
npt.assert_almost_equal(percept.data[34, 47, 0], np.max(percept.data))
# Full Argus II: 60 bright spots
model = ScoreboardModel(engine='serial', xystep=0.55, rho=100)
model.build()
percept = model.predict_percept(ArgusII(stim=np.ones(60)))
npt.assert_equal(np.sum(np.isclose(percept.data, 0.9, rtol=0.1, atol=0.1)),
60)
# Model gives same outcome as Spatial:
spatial = ScoreboardSpatial(engine='serial', xystep=1, rho=100)
spatial.build()
spatial_percept = model.predict_percept(ArgusII(stim=np.ones(60)))
npt.assert_almost_equal(percept.data, spatial_percept.data)
def test_ScoreboardModel():
# ScoreboardModel automatically sets `rho`:
model = ScoreboardModel(engine='serial', xystep=5)
npt.assert_equal(model.has_space, True)
npt.assert_equal(model.has_time, False)
npt.assert_equal(hasattr(model.spatial, 'rho'), True)
# User can set `rho`:
model.rho = 123
npt.assert_equal(model.rho, 123)
npt.assert_equal(model.spatial.rho, 123)
model.build(rho=987)
npt.assert_equal(model.rho, 987)
npt.assert_equal(model.spatial.rho, 987)
# Nothing in, None out:
npt.assert_equal(model.predict_percept(ArgusI()), None)
# Zero in = zero out:
implant = ArgusI(stim=np.zeros(16))
npt.assert_almost_equal(model.predict_percept(implant).data, 0)
# Multiple frames are processed independently:
model = ScoreboardModel(engine='serial', rho=200, xystep=5)
model.build()
percept = model.predict_percept(ArgusI(stim={'A1': [1, 2]}))
npt.assert_equal(percept.shape, list(model.grid.x.shape) + [2])
pmax = percept.data.max(axis=(0, 1))
npt.assert_almost_equal(percept.data[2, 3, :], pmax)
npt.assert_almost_equal(pmax[1] / pmax[0], 2.0)
def test_ProsthesisSystem_reshape_stim(rot, gtype, n_frames):
implant = ProsthesisSystem(ElectrodeGrid((10, 10), 30, rot=rot,
type=gtype))
# Smoke test the automatic reshaping:
n_px = 21
implant.stim = ImageStimulus(np.ones((n_px, n_px, n_frames)).squeeze())
npt.assert_equal(implant.stim.data.shape, (implant.n_electrodes, 1))
npt.assert_equal(implant.stim.time, None)
implant.stim = VideoStimulus(np.ones((n_px, n_px, 3 * n_frames)),
time=2 * np.arange(3 * n_frames))
npt.assert_equal(implant.stim.data.shape,
(implant.n_electrodes, 3 * n_frames))
npt.assert_equal(implant.stim.time, 2 * np.arange(3 * n_frames))
# Verify that a horizontal stimulus will always appear horizontally, even if
# the device is rotated:
data = np.zeros((50, 50))
data[20:-20, 10:-10] = 1
implant.stim = ImageStimulus(data)
model = ScoreboardModel(xrange=(-1, 1),
yrange=(-1, 1),
rho=30,
xystep=0.02)
model.build()
percept = label(model.predict_percept(implant).data.squeeze().T > 0.2)
npt.assert_almost_equal(regionprops(percept)[0].orientation, 0, decimal=1)
def test_deepcopy_ScoreboardModel():
original = ScoreboardModel()
copied = copy.deepcopy(original)
# Assert these are two different objects
npt.assert_equal(id(original) != id(copied), True)
# Assert these objects are equivalent
npt.assert_equal(original.__dict__ == copied.__dict__, True)
# Assert building one object does not affect the copied
original.build()
npt.assert_equal(copied.is_built, False)
npt.assert_equal(original.__dict__ != copied.__dict__, True)
# Assert destroying the original doesn't affect the copied
original = None
npt.assert_equal(copied is not None, True)
def test_ScoreboardModel():
# ScoreboardModel automatically sets `rho`:
model = ScoreboardModel(engine='serial', xystep=5)
npt.assert_equal(model.has_space, True)
npt.assert_equal(model.has_time, False)
npt.assert_equal(hasattr(model.spatial, 'rho'), True)
# User can set `rho`:
model.rho = 123
npt.assert_equal(model.rho, 123)
npt.assert_equal(model.spatial.rho, 123)
model.build(rho=987)
npt.assert_equal(model.rho, 987)
npt.assert_equal(model.spatial.rho, 987)
# Converting ret <=> dva
npt.assert_equal(isinstance(model.retinotopy, Watson2014Map), True)
npt.assert_almost_equal(model.retinotopy.ret2dva(0, 0), (0, 0))
npt.assert_almost_equal(model.retinotopy.dva2ret(0, 0), (0, 0))
model2 = ScoreboardModel(retinotopy=Watson2014DisplaceMap())
npt.assert_equal(isinstance(model2.retinotopy, Watson2014DisplaceMap),
True)
# Nothing in, None out:
npt.assert_equal(model.predict_percept(ArgusI()), None)
# Zero in = zero out:
implant = ArgusI(stim=np.zeros(16))
npt.assert_almost_equal(model.predict_percept(implant).data, 0)
# Multiple frames are processed independently:
model = ScoreboardModel(engine='serial',
rho=200,
xystep=5,
xrange=(-20, 20),
yrange=(-15, 15))
model.build()
percept = model.predict_percept(ArgusI(stim={'A1': [1, 2]}))
npt.assert_equal(percept.shape, list(model.grid.x.shape) + [2])
pmax = percept.data.max(axis=(0, 1))
npt.assert_almost_equal(percept.data[2, 3, :], pmax)
npt.assert_almost_equal(pmax[1] / pmax[0], 2.0)
npt.assert_almost_equal(percept.time, [0, 1])
def test_ScoreboardModel_predict_percept():
model = ScoreboardModel(xystep=0.55,
rho=100,
thresh_percept=0,
xrange=(-20, 20),
yrange=(-15, 15))
model.build()
# Single-electrode stim:
img_stim = np.zeros(60)
img_stim[47] = 1
percept = model.predict_percept(ArgusII(stim=img_stim))
# Single bright pixel, very small Gaussian kernel:
npt.assert_equal(np.sum(percept.data > 0.8), 1)
npt.assert_equal(np.sum(percept.data > 0.5), 2)
npt.assert_equal(np.sum(percept.data > 0.1), 7)
npt.assert_equal(np.sum(percept.data > 0.00001), 32)
# Brightest pixel is in lower right:
npt.assert_almost_equal(percept.data[33, 46, 0], np.max(percept.data))
# Full Argus II: 60 bright spots
model = ScoreboardModel(engine='serial', xystep=0.55, rho=100)
model.build()
percept = model.predict_percept(ArgusII(stim=np.ones(60)))
npt.assert_equal(np.sum(np.isclose(percept.data, 0.8, rtol=0.1, atol=0.1)),
88)
# Model gives same outcome as Spatial:
spatial = ScoreboardSpatial(engine='serial', xystep=1, rho=100)
spatial.build()
spatial_percept = model.predict_percept(ArgusII(stim=np.ones(60)))
npt.assert_almost_equal(percept.data, spatial_percept.data)
npt.assert_equal(percept.time, None)
# Warning for nonzero electrode-retina distances
implant = ArgusI(stim=np.ones(16), z=10)
msg = ("Nonzero electrode-retina distances do not have any effect on the "
"model output.")
assert_warns_msg(UserWarning, model.predict_percept, msg, implant)
# # .. _JobLib: https://joblib.readthedocs.io # .. _Dask: https://dask.org # # To change parameter values, either pass them directly to the constructor # above or set them by hand, like this: model.engine = 'serial' ############################################################################## # Then build the model. This is a necessary step before you can actually use # the model to predict a percept, as it performs a number of expensive setup # computations (e.g., building the spatial reference frame, calculating # electric potentials): model.build() ############################################################################## # .. note:: # # You need to build a model only once. After that, you can apply any number # of stimuli -- or even apply the model to different implants -- without # having to rebuild (which takes time). # # 2. Assigning a stimulus # ----------------------- # The second step is to specify a visual prosthesis from the # :py:mod:`~pulse2percept.implants` module. # # In the following, we will create an # :py:class:`~pulse2percept.implants.ArgusII` implant. By default, the implant