def grid_helper_plot_field(grid_helper, field, pylab):
"""
Plots the field
values
"""
field = field.copy()
# note transpose
zeros = field == 0
field[zeros] = np.nan
image = scale(field, min_value=0)
z = image[:, :, 0] # just R component
x = grid_helper._mgrids_plus[0]
y = grid_helper._mgrids_plus[1]
s0 = grid_helper._specs[0]
s1 = grid_helper._specs[1]
x = convert_unit(x, s0.units, s0.units_display)
y = convert_unit(y, s1.units, s1.units_display)
z = ma.masked_array(z, zeros)
pylab.pcolor(x, y, np.ones(z.shape), cmap="Pastel1")
pylab.pcolor(x, y, z, cmap="gray")
def write_similarity_matrix(A, out, more=False, images=None):
dtu.safe_pickle_dump(A, out + '.pickle')
# rgb = rgb_zoom(scale(A), 8)
rgb = scale(A)
rgb = dtu.d8_image_zoom_linear(rgb, 8)
dtu.write_image_as_jpg(rgb, out + '.jpg')
if more:
r = Report()
n = A.shape[0]
for i in range(n):
f = r.figure()
ignore = 10
Ai = A[i, :].copy()
for i2 in range(i - ignore, i + ignore):
if 0 <= i2 < n:
Ai[i2] = np.inf
jbest = np.argmin(Ai)
with f.plot('i-%d' % i) as pylab:
pylab.plot(A[i, :], '-*')
if images:
f.data_rgb(str(i), dtu.bgr_from_rgb(images[i]))
f.data_rgb(str(jbest), dtu.bgr_from_rgb(images[jbest]))
r.to_html(out + '.html')
def posneg_hinton(value, max_value=None, skim=0, nan_color=[0.5, 0.5, 0.5],
properties=None):
value = value.astype('float32')
# value = value.squeeze().copy() # squeeze: (1,1) -> ()
value = value.copy()
isfinite = np.isfinite(value)
isnan = np.logical_not(isfinite)
# set nan to 0
value[isnan] = 0
if max_value is None:
abs_value = abs(value)
if skim != 0:
abs_value = skim_top(abs_value, skim)
max_value = np.max(abs_value)
from reprep.graphics.filter_scale import scale
rgb_p = scale(value, min_value=0, max_value=max_value,
min_color=[0.5, 0.5, 0.5],
max_color=[1.0, 1.0, 1.0],
nan_color=[1, 0.6, 0.6],
flat_color=[0.5, 0.5, 0.5])
rgb_n = scale(value, min_value=-max_value, max_value=0,
max_color=[0.5, 0.5, 0.5],
min_color=[0.0, 0.0, 0.0],
nan_color=[1, 0.6, 0.6],
flat_color=[0.5, 0.5, 0.5])
w_p = value >= 0
# w_z = value == 0
w_n = value <= 0
H, W = value.shape
rgb = np.zeros((H, W, 3), 'uint8')
for i in range(3):
rgb[w_p, i] = rgb_p[w_p, i]
rgb[w_n, i] = rgb_n[w_n, i]
rgb[isnan, i] = nan_color[i]
# rgb[w_z, i] = 128
return rgb
def any_image_to_rgb(y):
def bail():
msg = 'Invalid image shape %s dtype %s' % (y.shape, y.dtype)
raise ValueError(msg)
if y.ndim == 2:
if not(y.dtype == 'float32'):
bail()
return scale(y)
if y.dtype == 'uint8':
if not y.ndim == 3 and y.shape[2] == 3:
bail()
return y
if not y.ndim == 3 and y.dtype == 'float32':
bail()
return (y * 255).astype('uint8')
def scalaruncertainty2rgb(x, umin=0, umax=1):
""" Converts the scalar uncertainty (in [min, max]) to rgb. (green=1, red=0) """
# set exactly 0 to nan, and color it gray
x = x.copy()
# if min is None:
# min = np.min(x)
# if max is None:
# max = np.nanmax(x)
x[x == 0] = np.nan
rgb = scale(x, max_value=umax, min_value=umin,
min_color=[1, 0, 0], max_color=[0, 1, 0],
nan_color=[0.5, 0.5, 0.5])
return rgb
def any_image_to_rgb(y):
def bail():
msg = 'Invalid image shape %s dtype %s' % (y.shape, y.dtype)
raise ValueError(msg)
if y.ndim == 2:
if not (y.dtype == 'float32'):
bail()
return scale(y)
if y.dtype == 'uint8':
if not y.ndim == 3 and y.shape[2] == 3:
bail()
return y
if not y.ndim == 3 and y.dtype == 'float32':
bail()
return (y * 255).astype('uint8')
def plot_phi_d_diagram_bgr(lane_filter, belief, phi, d, dpi=120,
bgcolor=dtu.ColorConstants.RGB_DUCKIETOWN_YELLOW,
other_phi=None, other_d=None):
""" Returns a BGR image """
figure_args = dict(facecolor=dtu.matplotlib_01_from_rgb(bgcolor))
a = dtu.CreateImageFromPylab(dpi=dpi, figure_args=figure_args)
with a as pylab:
f_d = lambda x: 100 * x
f_phi = np.rad2deg
# Where are the lanes?
lane_width = lane_filter.lanewidth
d_max = lane_filter.d_max
d_min = lane_filter.d_min
phi_max = lane_filter.phi_max
phi_min = lane_filter.phi_min
delta_d = lane_filter.delta_d
delta_phi = lane_filter.delta_phi
# note transpose
belief = belief.copy()
zeros = belief == 0
belief[zeros] = np.nan
belief_image = scale(belief, min_value=0)
x = f_d(lane_filter.d_pcolor)
y = f_phi(lane_filter.phi_pcolor)
z = belief_image[:, :, 0] # just R component
z = ma.masked_array(z, zeros)
pylab.pcolor(x, y, np.ones(z.shape), cmap='Pastel1')
pylab.pcolor(x, y, z, cmap='gray')
if other_phi is not None:
for _phi, _d in zip(other_phi, other_d):
pylab.plot(f_d(_d), f_phi(_phi), 'go', markersize=15,
markeredgecolor='none',
markeredgewidth=3,
markerfacecolor='blue')
pylab.plot(f_d(d), f_phi(phi), 'go', markersize=20,
markeredgecolor='magenta',
markeredgewidth=3,
markerfacecolor='none')
pylab.plot(f_d(d), f_phi(phi), 'o', markersize=2,
markeredgecolor='none',
markeredgewidth=0,
markerfacecolor='magenta')
W = f_d(lane_width / 2)
width_white = f_d(lane_filter.linewidth_white)
width_yellow = f_d(lane_filter.linewidth_yellow)
pylab.plot([-W, -W], [f_phi(phi_min), f_phi(phi_max)], 'k-')
pylab.plot([-W - width_white, -W - width_white], [f_phi(phi_min), f_phi(phi_max)], 'k-')
pylab.plot([0, 0], [f_phi(phi_min), f_phi(phi_max)], 'k-')
pylab.plot([+W, +W], [f_phi(phi_min), f_phi(phi_max)], 'y-')
pylab.plot([+W + width_yellow, +W + width_yellow], [f_phi(phi_min), f_phi(phi_max)], 'y-')
s = ''
s += "status = %s" % lane_filter.getStatus()
s += "\nphi = %.1f deg" % f_phi(phi)
s += "\nd = %.1f cm" % f_d(d)
s += "\nentropy = %.4f" % lane_filter.get_entropy()
s += "\nmax = %.4f" % belief.max()
s += "\nmin = %.4f" % belief.min()
if other_phi is not None:
s += '\n Other answers:'
for _phi, _d in zip(other_phi, other_d):
s += "\nphi = %.1f deg" % f_phi(_phi)
s += "\nd = %.1f cm" % f_d(_d)
y = f_phi(phi_max) - 10
args = dict(rotation=-90, color='white')
annotate = True
if annotate:
pylab.annotate(s, xy=(0.7, 0.35), xycoords='figure fraction')
pylab.annotate("in middle of right lane", xy=(0, y), **args)
pylab.annotate("on right white tape", xy=(-W, y), **args)
pylab.annotate("on left yellow tape", xy=(+W, y), **args)
pylab.annotate("in other lane", xy=(+W * 1.3, y), **args)
pylab.axis([f_d(d_min), f_d(d_max), f_phi(phi_min), f_phi(phi_max)])
pylab.ylabel('phi: orientation (deg); cell = %.1f deg' % f_phi(delta_phi))
pylab.xlabel('d: distance from center line (cm); cell = %.1f cm' % f_d(delta_d))
return a.get_bgr()
def plot_phi_d_diagram_bgr(
lane_filter,
belief,
phi,
d,
dpi: int = 120,
bgcolor: dtu.BGRColor8 = dtu.ColorConstants.BGR_DUCKIETOWN_YELLOW,
other_phi=None,
other_d=None,
) -> dtu.NPImageBGR:
"""Returns a BGR image"""
facecolor = dtu.matplotlib_01_from_rgb(
dtu.rgb_color_from_bgr_color(bgcolor))
figure_args = dict(facecolor=facecolor)
a = dtu.CreateImageFromPylab(dpi=dpi, figure_args=figure_args)
with a as pylab:
f_d = lambda x: 100 * x
f_phi = np.rad2deg
# Where are the lanes?
lane_width = lane_filter.lanewidth
d_max = lane_filter.d_max
d_min = lane_filter.d_min
phi_max = lane_filter.phi_max
phi_min = lane_filter.phi_min
delta_d = lane_filter.delta_d
delta_phi = lane_filter.delta_phi
# note transpose
belief = belief.copy()
zeros = belief == 0
belief[zeros] = np.nan
belief_image = scale(belief, min_value=0)
x = f_d(lane_filter.d_pcolor)
y = f_phi(lane_filter.phi_pcolor)
z = belief_image[:, :, 0] # just R component
z = ma.masked_array(z, zeros)
pylab.pcolor(x, y, np.ones(z.shape), cmap="Pastel1")
pylab.pcolor(x, y, z, cmap="gray")
if other_phi is not None:
for _phi, _d in zip(other_phi, other_d):
pylab.plot(
f_d(_d),
f_phi(_phi),
"go",
markersize=15,
markeredgecolor="none",
markeredgewidth=3,
markerfacecolor="blue",
)
pylab.plot(
f_d(d),
f_phi(phi),
"go",
markersize=20,
markeredgecolor="magenta",
markeredgewidth=3,
markerfacecolor="none",
)
pylab.plot(
f_d(d),
f_phi(phi),
"o",
markersize=2,
markeredgecolor="none",
markeredgewidth=0,
markerfacecolor="magenta",
)
W = f_d(lane_width / 2)
width_white = f_d(lane_filter.linewidth_white)
width_yellow = f_d(lane_filter.linewidth_yellow)
pylab.plot([-W, -W], [f_phi(phi_min), f_phi(phi_max)], "k-")
pylab.plot([-W - width_white, -W - width_white],
[f_phi(phi_min), f_phi(phi_max)], "k-")
pylab.plot([0, 0], [f_phi(phi_min), f_phi(phi_max)], "k-")
pylab.plot([+W, +W], [f_phi(phi_min), f_phi(phi_max)], "y-")
pylab.plot([+W + width_yellow, +W + width_yellow],
[f_phi(phi_min), f_phi(phi_max)], "y-")
s = ""
s += f"status = {lane_filter.getStatus()}"
s += f"\nphi = {f_phi(phi):.1f} deg"
s += f"\nd = {f_d(d):.1f} cm"
s += f"\nentropy = {lane_filter.get_entropy():.4f}"
s += f"\nmax = {belief.max():.4f}"
s += f"\nmin = {belief.min():.4f}"
if other_phi is not None:
s += "\n Other answers:"
for _phi, _d in zip(other_phi, other_d):
s += f"\nphi = {f_phi(_phi):.1f} deg"
s += f"\nd = {f_d(_d):.1f} cm"
y = f_phi(phi_max) - 10
args = dict(rotation=-90, color="white")
annotate = True
if annotate:
pylab.annotate(s, xy=(0.7, 0.35), xycoords="figure fraction")
pylab.annotate("in middle of right lane", xy=(0, y), **args)
pylab.annotate("on right white tape", xy=(-W, y), **args)
pylab.annotate("on left yellow tape", xy=(+W, y), **args)
pylab.annotate("in other lane", xy=(+W * 1.3, y), **args)
pylab.axis([f_d(d_min), f_d(d_max), f_phi(phi_min), f_phi(phi_max)])
pylab.ylabel(
f"phi: orientation (deg); cell = {f_phi(delta_phi):.1f} deg")
pylab.xlabel(
f"d: distance from center line (cm); cell = {f_d(delta_d):.1f} cm")
return a.get_bgr()
def diffeo_show_mismatch(y0, y1):
e = np.abs(y0 - y1)
if e.ndim == 3:
e.sum(axis=2)
return scale(e, min_color=[0, 0, 0], max_color=[1, 0, 0])
def diffeo_show_mismatch(y0, y1):
e = np.abs(y0 - y1)
if e.ndim == 3:
e.sum(axis=2)
return scale(e, min_color=[0, 0, 0], max_color=[1, 0, 0])
