def test_predict_batched(engine):
if not has_jax:
pytest.skip("Jax not installed")
# Allows mix of valid Stimulus types
stims = [{
'A5': BiphasicPulseTrain(25, 4, 0.45),
'C7': BiphasicPulseTrain(50, 2.5, 0.75)
}, {
'B4': BiphasicPulseTrain(3, 1, 0.32)
},
Stimulus({'F3': BiphasicPulseTrain(12, 3, 1.2)})]
implant = ArgusII()
model = BiphasicAxonMapModel(engine=engine, xystep=2)
model.build()
# Import error if we dont have jax
if engine != 'jax':
with pytest.raises(ImportError):
model.predict_percept_batched(implant, stims)
return
percepts_batched = model.predict_percept_batched(implant, stims)
percepts_serial = []
for stim in stims:
implant.stim = stim
percepts_serial.append(model.predict_percept(implant))
npt.assert_equal(len(percepts_serial), len(percepts_batched))
for p1, p2 in zip(percepts_batched, percepts_serial):
npt.assert_almost_equal(p1.data, p2.data)
def test_AxonMapModel(engine):
set_params = {'xystep': 2, 'engine': engine, 'rho': 432, 'axlambda': 2,
'n_axons': 9, 'n_ax_segments': 50,
'xrange': (-30, 30), 'yrange': (-20, 20),
'loc_od_x': 5, 'loc_od_y': 6}
model = AxonMapModel()
for param in set_params:
npt.assert_equal(hasattr(model.spatial, param), True)
# User can override default values
for key, value in set_params.items():
setattr(model.spatial, key, value)
npt.assert_equal(getattr(model.spatial, key), value)
model = AxonMapModel(**set_params)
model.build(**set_params)
for key, value in set_params.items():
npt.assert_equal(getattr(model.spatial, key), value)
# Zeros in, zeros out:
implant = ArgusII(stim=np.zeros(60))
npt.assert_almost_equal(model.predict_percept(implant).data, 0)
implant.stim = np.zeros(60)
npt.assert_almost_equal(model.predict_percept(implant).data, 0)
# Implant and model must be built for same eye:
with pytest.raises(ValueError):
implant = ArgusII(eye='LE', stim=np.zeros(60))
model.predict_percept(implant)
with pytest.raises(ValueError):
AxonMapModel(eye='invalid').build()
with pytest.raises(ValueError):
AxonMapModel(xystep=5).build(eye='invalid')
def test_predict_spatial_jax():
# ensure jax predict spatial is equal to normal
if not has_jax:
pytest.skip("Jax not installed")
model1 = BiphasicAxonMapModel(engine='jax', xystep=2)
model2 = BiphasicAxonMapModel(engine='cython', xystep=2)
model1.build()
model2.build()
implant = ArgusII()
implant.stim = {
'A5': BiphasicPulseTrain(25, 4, 0.45),
'C7': BiphasicPulseTrain(50, 2.5, 0.75)
}
p1 = model1.predict_percept(implant)
p2 = model2.predict_percept(implant)
npt.assert_almost_equal(p1.data, p2.data, decimal=4)
# test changing model parameters, make sure jax is clearing cache on build
model1.axlambda = 800
model2.axlambda = 800
model1.rho = 50
model2.rho = 50
model1.build()
model2.build()
p1 = model1.predict_percept(implant)
p2 = model2.predict_percept(implant)
npt.assert_almost_equal(p1.data, p2.data, decimal=4)
def test_AxonMapModel(engine):
set_params = {
'xystep': 2,
'engine': engine,
'rho': 432,
'axlambda': 20,
'n_axons': 9,
'n_ax_segments': 50,
'xrange': (-30, 30),
'yrange': (-20, 20),
'loc_od': (5, 6)
}
model = AxonMapModel()
for param in set_params:
npt.assert_equal(hasattr(model.spatial, param), True)
# User can override default values
for key, value in set_params.items():
setattr(model.spatial, key, value)
npt.assert_equal(getattr(model.spatial, key), value)
model = AxonMapModel(**set_params)
model.build(**set_params)
for key, value in set_params.items():
npt.assert_equal(getattr(model.spatial, key), value)
# 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 = AxonMapModel(retinotopy=Watson2014DisplaceMap())
npt.assert_equal(isinstance(model2.retinotopy, Watson2014DisplaceMap),
True)
# Zeros in, zeros out:
implant = ArgusII(stim=np.zeros(60))
npt.assert_almost_equal(model.predict_percept(implant).data, 0)
implant.stim = np.zeros(60)
npt.assert_almost_equal(model.predict_percept(implant).data, 0)
# Implant and model must be built for same eye:
with pytest.raises(ValueError):
implant = ArgusII(eye='LE', stim=np.zeros(60))
model.predict_percept(implant)
with pytest.raises(ValueError):
AxonMapModel(eye='invalid').build()
with pytest.raises(ValueError):
AxonMapModel(xystep=5).build(eye='invalid')
# Lambda cannot be too small:
with pytest.raises(ValueError):
AxonMapModel(axlambda=9).build()
# will be centered over the fovea (at x=0, y=0) and aligned with the horizontal # meridian (rot=0): from pulse2percept.implants import ArgusII implant = ArgusII() ############################################################################## # The easiest way to assign a stimulus to the implant is to pass a NumPy array # that specifies the current amplitude to be applied to every electrode in the # implant. # # For example, the following sends 10 microamps to all 60 electrodes of the # implant: import numpy as np implant.stim = 10 * np.ones(60) ############################################################################## # .. note:: # # Some models can handle stimuli that have both a spatial and a temporal # component. the scoreboard model cannot. # # 3. Predicting the percept # ------------------------- # The third step is to apply the model to predict the percept resulting from # the specified stimulus. Note that this may take some time on your machine: percept = model.predict_percept(implant) ##############################################################################
model.plot() implant.plot() ############################################################################## # By default, the plots will be added to the current Axes object. # Alternatively, you can pass ``ax=`` to specify in which Axes to plot. # # The easiest way to assign a stimulus to the implant is to pass a NumPy array # that specifies the current amplitude to be applied to every electrode in the # implant. # # For example, the following sends 1 microamp to all 60 electrodes of the # implant: import numpy as np implant.stim = np.ones(60) ############################################################################## # Predicting the percept # ---------------------- # The third step is to apply the model to predict the percept resulting from # the specified stimulus. Note that this may take some time on your machine: percept = model.predict_percept(implant) ############################################################################## # The resulting percept is stored in a # :py:class:`~pulse2percept.percepts.Percept` object, which is similar in # organization to the :py:class:`~pulse2percept.stimuli.Stimulus` object: # the ``data`` container is a 3D NumPy array (Y, X, T) with labeled axes # ``xdva``, ``ydva``, and ``time``.
def test_biphasicAxonMapSpatial(engine):
# Lambda cannot be too small:
with pytest.raises(ValueError):
BiphasicAxonMapSpatial(axlambda=9).build()
model = BiphasicAxonMapModel(engine=engine, xystep=2).build()
# Jax not implemented yet
if engine == 'jax':
with pytest.raises(NotImplementedError):
implant = ArgusII()
implant.stim = Stimulus({'A5': BiphasicPulseTrain(20, 1, 0.45)})
percept = model.predict_percept(implant)
return
# Only accepts biphasic pulse trains with no delay dur
implant = ArgusI(stim=np.ones(16))
with pytest.raises(TypeError):
model.predict_percept(implant)
# Nothing in, None out:
npt.assert_equal(model.predict_percept(ArgusI()), None)
# Zero in = zero out:
implant = ArgusI(stim=np.zeros(16))
percept = model.predict_percept(implant)
npt.assert_equal(isinstance(percept, Percept), True)
npt.assert_equal(percept.shape, list(model.grid.x.shape) + [1])
npt.assert_almost_equal(percept.data, 0)
npt.assert_equal(percept.time, None)
# Should be equal to axon map model if effects models return 1
model = BiphasicAxonMapSpatial(engine=engine, xystep=2)
def bright_model(freq, amp, pdur):
return 1
def size_model(freq, amp, pdur):
return 1
def streak_model(freq, amp, pdur):
return 1
model.bright_model = bright_model
model.size_model = size_model
model.streak_model = streak_model
model.build()
axon_map = AxonMapSpatial(xystep=2).build()
implant = ArgusII()
implant.stim = Stimulus({'A5': BiphasicPulseTrain(20, 1, 0.45)})
percept = model.predict_percept(implant)
percept_axon = axon_map.predict_percept(implant)
npt.assert_almost_equal(percept.data[:, :, 0],
percept_axon.get_brightest_frame())
# Effect models must be callable
model = BiphasicAxonMapSpatial(engine=engine, xystep=2)
model.bright_model = 1.0
with pytest.raises(TypeError):
model.build()
# If t_percept is not specified, there should only be one frame
model = BiphasicAxonMapSpatial(engine=engine, xystep=2)
model.build()
implant = ArgusII()
implant.stim = Stimulus({'A5': BiphasicPulseTrain(20, 1, 0.45)})
percept = model.predict_percept(implant)
npt.assert_equal(percept.time is None, True)
# If t_percept is specified, only first frame should have data
# and the rest should be empty
percept = model.predict_percept(implant, t_percept=[0, 1, 2, 5, 10])
npt.assert_equal(len(percept.time), 5)
npt.assert_equal(np.any(percept.data[:, :, 0]), True)
npt.assert_equal(np.any(percept.data[:, :, 1:]), False)
# Test that default models give expected values
model = BiphasicAxonMapSpatial(engine=engine,
rho=400,
axlambda=600,
xystep=1,
xrange=(-20, 20),
yrange=(-15, 15))
model.build()
implant = ArgusII()
implant.stim = Stimulus({'A4': BiphasicPulseTrain(20, 1, 1)})
percept = model.predict_percept(implant)
npt.assert_equal(np.sum(percept.data > 0.0813), 81)
npt.assert_equal(np.sum(percept.data > 0.1626), 59)
npt.assert_equal(np.sum(percept.data > 0.2439), 44)
npt.assert_equal(np.sum(percept.data > 0.4065), 26)
npt.assert_equal(np.sum(percept.data > 0.5691), 14)
def test_biphasicAxonMapModel(engine):
set_params = {
'xystep': 2,
'engine': engine,
'rho': 432,
'axlambda': 20,
'n_axons': 9,
'n_ax_segments': 50,
'xrange': (-30, 30),
'yrange': (-20, 20),
'loc_od': (5, 6),
'do_thresholding': False
}
model = BiphasicAxonMapModel(engine=engine)
for param in set_params:
npt.assert_equal(hasattr(model.spatial, param), True)
# We can set and get effects model params
for atr in ['a' + str(i) for i in range(0, 10)]:
npt.assert_equal(hasattr(model, atr), True)
model.a0 = 5
# Should propogate to size and bright model
# But should not be a member of streak or spatial
npt.assert_equal(model.spatial.size_model.a0, 5)
npt.assert_equal(model.spatial.bright_model.a0, 5)
npt.assert_equal(hasattr(model.spatial.streak_model, 'a0'), False)
with pytest.raises(AttributeError):
model.spatial.__getattribute__('a0')
# If the spatial model and an effects model have a parameter with the
# Same name, both need to be changed
model.rho = 350
model.axlambda = 450
model.do_thresholding = True
npt.assert_equal(model.spatial.size_model.rho, 350)
npt.assert_equal(model.spatial.streak_model.axlambda, 450)
npt.assert_equal(model.spatial.bright_model.do_thresholding, True)
npt.assert_equal(model.rho, 350)
npt.assert_equal(model.axlambda, 450)
npt.assert_equal(model.do_thresholding, True)
# Effect model parameters can be passed even in constructor
model = BiphasicAxonMapModel(engine=engine, a0=5, rho=432)
npt.assert_equal(model.a0, 5)
npt.assert_equal(model.spatial.bright_model.a0, 5)
npt.assert_equal(model.rho, 432)
npt.assert_equal(model.spatial.size_model.rho, 432)
# If parameter is not an effects model param, it cant be set
with pytest.raises(FreezeError):
model.invalid_param = 5
# Custom parameters also propogate to effects models
model = BiphasicAxonMapModel(engine=engine)
class TestSizeModel():
def __init__(self):
self.test_param = 5
def __call__(self, freq, amp, pdur):
return 1
model.size_model = TestSizeModel()
model.test_param = 10
npt.assert_equal(model.spatial.size_model.test_param, 10)
with pytest.raises(AttributeError):
model.spatial.__getattribute__('test_param')
# Values are passed correctly even in another classes __init__
# This also tests for recursion error in another classes __init__
class TestInitClassGood():
def __init__(self):
self.model = BiphasicAxonMapModel()
# This shouldnt raise an error
self.model.a0
class TestInitClassBad():
def __init__(self):
self.model = BiphasicAxonMapModel()
# This should
self.model.a10 = 999
# If this fails, something is wrong with getattr / setattr logic
TestInitClassGood()
with pytest.raises(FreezeError):
TestInitClassBad()
# User can override default values
model = BiphasicAxonMapModel(engine=engine)
for key, value in set_params.items():
setattr(model.spatial, key, value)
npt.assert_equal(getattr(model.spatial, key), value)
model = BiphasicAxonMapModel(**set_params)
model.build(**set_params)
for key, value in set_params.items():
npt.assert_equal(getattr(model.spatial, key), value)
# Zeros in, zeros out:
implant = ArgusII(stim=np.zeros(60))
npt.assert_almost_equal(model.predict_percept(implant).data, 0)
implant.stim = np.zeros(60)
npt.assert_almost_equal(model.predict_percept(implant).data, 0)
# Implant and model must be built for same eye:
with pytest.raises(ValueError):
implant = ArgusII(eye='LE', stim=np.zeros(60))
model.predict_percept(implant)
with pytest.raises(ValueError):
BiphasicAxonMapModel(eye='invalid').build()
with pytest.raises(ValueError):
BiphasicAxonMapModel(xystep=5).build(eye='invalid')
# Lambda cannot be too small:
with pytest.raises(ValueError):
BiphasicAxonMapModel(axlambda=9).build()
# :py:class:`~pulse2percept.implants.ArgusII` implant and use the # :py:class:`~pulse2percept.models.AxonMapModel` [Beyeler2019]_ to interpret it: # # .. important :: # # Don't forget to build the model before using ``predict_percept`` # from pulse2percept.implants import ArgusII from pulse2percept.models import AxonMapModel model = AxonMapModel() model.build() implant = ArgusII() implant.stim = GratingStimulus((25, 25), temporal_freq=0.1) percept = model.predict_percept(implant) percept.play() ##################################################################################### # As you can see in the above code segment, the stimulus passed to the implant does # not necessarily have to have the same dimensions as the electrode grid. # This is functionality built in to the implant code: The implant will automatically # rescale the stimulus to the appropriate size. # In the case of Argus II, the stimulus would thus be downscaled to a 6x10 image. # # Pre-Processing Stimuli # ---------------------- # # Since both :py:class:`~pulse2percept.stimuli.BarStimulus` and
############################################################################### # Now we need to activate one electrode at a time, and predict the resulting # percept. We could build a :py:class:`~pulse2percept.stimuli.Stimulus` object # with a for loop that does just that, or we can use the following trick. # # The stimulus' data container is a (electrodes, timepoints) shaped 2D NumPy # array. Activating one electrode at a time is therefore the same as an # identity matrix whose size is equal to the number of electrodes. In code: # Find the names of all the electrodes in the dataset: electrodes = data.electrode.unique() # Activate one electrode at a time: import numpy as np from pulse2percept.stimuli import Stimulus argus.stim = Stimulus(np.eye(len(electrodes)), electrodes=electrodes) ############################################################################### # Using the model's # :py:func:`~pulse2percept.models.AxonMapModel.predict_percept`, we then get # a Percept object where each frame is the percept generated from activating # a single electrode: percepts = model.predict_percept(argus) percepts.play() ############################################################################### # Finally, we can visualize the ground-truth and simulated phosphenes # side-by-side: from pulse2percept.viz import plot_argus_simulated_phosphenes
# will be centered over the fovea (at x=0, y=0) and aligned with the horizontal # meridian (rot=0): from pulse2percept.implants import ArgusII implant = ArgusII() ############################################################################## # The easiest way to assign a stimulus to the implant is to pass a NumPy array # that specifies the current amplitude to be applied to every electrode in the # implant. # # For example, the following sends 10 microamps to all 60 electrodes of the # implant: import numpy as np implant.stim = 10 * np.ones(60) ############################################################################## # .. note:: # # Some models can handle stimuli that have both a spatial and a temporal # component. the scoreboard model cannot. # # Predicting the percept # ---------------------- # The third step is to apply the model to predict the percept resulting from # the specified stimulus. Note that this may take some time on your machine: percept = model.predict_percept(implant) ##############################################################################
implant.plot()
plt.show()
##############################################################################
# As mentioned above, the Biphasic Axon Map Model only accepts
# :py:class:`~pulse2percept.stimuli.BiphasicPulseTrain`
# stimuli with no :py:attr:`~pulse2percept.stimuli.BiphasicPulseTrain.delay_dur`.
# The amplitude given to the BiphasicPulseTrain
# is interpreted as amplitude as a factor of threshold (i.e. an amp of 1 means
# 1xTh)
#
# You can easily assign BiphasicPulseTrains to electrodes with a dictionary
# The following creates a train with 20Hz frequency, 1xTh amplitude, and 0.45ms
# pulse / phase duration.
implant.stim = {'A4': BiphasicPulseTrain(20, 1, 0.45)}
implant.stim.plot()
##############################################################################
# Finally, you can predict the percept resulting from stimulation
percept = model.predict_percept(implant)
ax = percept.plot()
ax.set_title('Predicted percept')
plt.show()
##############################################################################
# Increasing the frequency will make phosphenes brighter
fig, axes = plt.subplots(1, 2, sharex=True, sharey=True)
implant.stim = {'A4': BiphasicPulseTrain(50, 1, 0.45)}
new_percept = model.predict_percept(implant)
new_percept.plot(ax=axes[1])