def generate_whitenoise_sequence(seq_array_bool,
seq_array,
scale: int,
debug=False,
mask=None,
**kwargs):
""" Generates a sequence for upload to DMD.
:param seq_array_bool: boolean ndarray to write the random bits, dimensions (N_pix, H_dmd/scale, W_dmd/scale)
:param seq_array: uint8 ndarray for upload with dimensions (N_pix, H_dmd, W_dmd)
:param scale: scale for discreet random pixels in DMD space. ie if scale=2, each random pixel will be
projected as 2x2 pix on the dmd.
:param debug: use numeric debug sequence generator to test synchronization.
"""
if not debug:
random_sequence_generator(seq_array_bool)
zoomer(seq_array_bool, scale, seq_array)
else:
global _DEBUG_COUNTER # a bit messy, but I don't want to spend to much time on the debug cleanliness.
# _DEBUG_COUNTER += seq_array_bool.shape[0]
number_sequence_generator(_DEBUG_COUNTER, seq_array)
seq_array *= 255
_DEBUG_COUNTER += seq_array_bool.shape[0]
# print(seq_array.mean())
seq_array[:] *= mask # mask is 1 in areas we want to stimulate and 0 otherwise. This is faster than alternatives.
def sequence_generator(seq_array_bool,
seq_array,
scale: int,
debug=False,
mask=None):
""" Generates a sequence for upload to DMD.
:param seq_array_bool: boolean ndarray to write the random bits, dimensions (N_pix, H_dmd/scale, W_dmd/scale)
:param seq_array: uint8 ndarray for upload with dimensions (N_pix, H_dmd, W_dmd)
:param scale: scale for discreet random pixels in DMD space. ie if scale=2, each random pixel will be
projected as 2x2 pix on the dmd.
:param debug: use numeric debug sequence generator to test synchronization.
"""
if random_threshold is None:
raise ValueError('random_threshold kwarg is not set.')
unmasked = find_unmasked_px(mask, scale)
n_nmasked = unmasked.sum()
n_frames, h, w = seq_array_bool.shape
total_randnums = n_nmasked * n_frames
randnums = np.random.rand(total_randnums)
randbool = randnums <= random_threshold
reshape(randbool, unmasked, seq_array_bool)
zoomer(seq_array_bool, scale, seq_array)
seq_array[:] *= mask # mask is 1 in areas we want to stimulate and 0 otherwise. This is faster than alternatives.
def make_patterns(self, boolean_array: np.ndarray,
whole_seq_array: np.ndarray, debug):
"""
Modifies arrays in place with random values (on, off).
This makes sparse random patterns drawn from varying probability distributions.
:param boolean_array: boolean array that is of shape ( n_frames, h / scale, w / scale)
:param whole_seq_array: array of uint8 values of shave (n_frames, h, w)
:param debug: not implemented.
"""
n_frames, h, w = boolean_array.shape
total_randnums = n_frames * self.n_unmasked_pix
probs = np.zeros(total_randnums)
self._frame_count = self._gen_probs(probs, self._frame_count,
self.n_unmasked_pix,
self.switch_frequency,
self.random_probs, self._last_p)
self._last_p = probs[
-1] # save this for the next iteration if we're in the middle of a presentation block.
a = np.random.binomial(1, probs, total_randnums)
utils.reshape(a, self.unmasked, boolean_array)
utils.zoomer(boolean_array, self.scale, whole_seq_array)
whole_seq_array *= self.mask
def make_patterns(self, boolean_array, whole_seq_array, debug):
n_frames, h, w = boolean_array.shape
_, h_full, w_full = whole_seq_array.shape
assert h_full // h == self.scale
# whole_seq_array[:, :, :] = 0
boolean_array[:, :, :] = False
for i in range(n_frames):
msg = '{:06n}'.format(self._count)
# self._make_text_fast(msg, whole_seq_array[i, :, :])
self._make_text_fast(msg, boolean_array[i, :, :], margins=(2, 2, 37, 37))
self._count += 1
utils.zoomer(boolean_array, self.scale, whole_seq_array)
def make_patterns(self, boolean_array: np.ndarray,
whole_seq_array: np.ndarray, debug):
"""
Modifies arrays in place with random values (on, off).
:param boolean_array: boolean array that is of shape ( n_frames, h / scale, w / scale)
:param whole_seq_array: array of uint8 values of shave (n_frames, h, w)
:param debug: not implemented.
:return:
"""
n_frames, h, w = boolean_array.shape
total_randnums = n_frames * self.n_unmasked_pix
randnums = np.random.rand(total_randnums)
randbool = randnums <= self.threshold
utils.reshape(randbool, self.unmasked, boolean_array)
utils.zoomer(boolean_array, self.scale, whole_seq_array)
whole_seq_array *= self.mask