def rotated_mean(N=100):
images = lambda: util.images().pad(100, 100)
angles = lambda: util.images().pad(100, 100).map(transform.PCA_angle)
rotated = lambda: util.seq(transform.rotate(img, ang) for img, ang in zip(images(), angles()))
logger.info("Calculating mean rotated image")
mean_r = util.mean(rotated(), np.zeros((100, 100)))
# distance
d = np.linalg.norm
dist = (d(img - mean_r) for img in util.progress(rotated()))
logger.info(f"Finding {N} farthest away from the mean")
oddities = nlargest(N, zip(dist, util.image_ids()))
logger.info(f"Finished rotated mean outliers")
return oddities
def centered_mean(N=100):
centered = lambda: util.images().pad(100, 100).map(transform.center)
logger.info("Calculating mean centered image")
mean_c = util.mean(centered(), np.zeros((100, 100)))
# distance
d = np.linalg.norm
dist = (d(img - mean_c) for img in util.progress(centered()))
logger.info(f"Finding {N} farthest away from the mean")
oddities = nlargest(N, zip(dist, util.image_ids()))
logger.info(f"Finished centered mean outliers")
return oddities
def amplitude_mean(N=100):
amps = lambda: util.images().pad(100, 100).map(transform.amplitude)
logger.info("Calculating mean amplitude")
mean_a = util.mean(amps(), np.zeros((100, 100)))
# distance
d = np.linalg.norm
dist = (d(amp - mean_a) for amp in util.progress(amps()))
logger.info(f"Finding {N} farthest away from the mean")
oddities = nlargest(N, zip(dist, util.image_ids()))
logger.info(f"Finished amplitude mean outliers")
return oddities
def mean(N=100):
images = lambda: util.images().pad(100, 100)
logger.info("Calculating mean image")
mean = util.mean(images())
# max-correlation of mean
d = util.max_correlation
d = np.linalg.norm
maxcorr = (d(img - mean) for img in util.progress(images()))
logger.info(f"Finding {N} with least correlation with the mean")
oddities = nlargest(N, zip(maxcorr, util.image_ids()))
logger.info(f"Finished maxcorr outliers")
return oddities
def histogram_mean(N=100):
hists = lambda: util.images().map(transform.smoothen).map(transform.grayscale)
logger.info("Calculating mean histogram")
mean_h = util.mean(hists(), np.zeros((256)))
# distance
d = np.linalg.norm
d = util.max_correlation1d
dist = (d(hist, mean_h) for hist in util.progress(hists()))
logger.info(f"Finding {N} farthest away from the mean")
oddities = nlargest(N, zip(dist, util.image_ids()))
logger.info(f"Finished histogram mean outliers")
return oddities
def frequency_mean(N=100):
phases = lambda: util.images().pad(100, 100).map(transform.phase)
logger.info("Calculating mean phase")
mean_p = util.mean(phases(), np.zeros((100, 100)))
# mean_p = transform.shape(mean_p, shape="circle", mask_value=0, radius=25)
# distance
d = np.linalg.norm
d = util.max_correlation
dist = (-d(phase, mean_p) for phase in util.progress(phases()))
logger.info(f"Finding {N} farthest away from the mean")
oddities = nlargest(N, zip(dist, util.image_ids()))
logger.info(f"Finished phase mean outliers")
return oddities
def PCA_angle(N=100):
"""
Computes the mean angle, the finds N angles that deviates the most from
the mean, modulo pi.
There was an issue with the angle of the PCA to be inverted 180 degrees.
This shouldn't affect the mean too much, but caused some images to be
incorrectly classified as anomalies. Thus, we consider only angles
between [-90, 90).
"""
angles = lambda: util.images().pad(100, 100).map(transform.PCA_angle)
logger.info("Calculating PCA angle mean")
mean = util.angle_mean(angles())
# distances from mean angle
distances = (util.circle_dist(mean, angle) for angle in angles())
logger.info(f"Finding {N} angles farthest away from mean")
oddities = nlargest(N, zip(distances, util.image_ids()))
logger.info(f"Finished PCA angle outliers")
return oddities
for k, (i, j) in enumerate(block_loop(n)):
axes[i, j].plot(circle_x, circle_y, color='k', lw=1.2)
axes[i, j].plot(*line(angles[k]), color='k', lw=1.2)
with Figure(2, 5) as fig:
axes = get_axes(2, 5, fig)
arrow(axes[..., 2])
for col_index in [0, 1, 3, 4]:
clean(axes[..., col_index])
square(axes[..., col_index])
angles = lambda: (util.images().pad().map(transform.PCA_angle))
# block 1
img_mean = util.mean(util.images().pad(100, 100))
img_3 = [
util.pad(util.image(i), 100, 100) for i in ["1377", "1590", "9728"]
]
samples = [img_mean] + img_3
block_imshow(1, axes, samples)
# block 2
angle_mean = util.angle_mean(angles())
angle_3 = list(map(transform.PCA_angle, img_3))
img = util.DUMMY_IMAGE()
img = center(img)
ax.imshow(img, cmap="gray")
fig.update(1)
#--- Fixed shift
if False:
with Figure(1, 1) as fig:
ax = fig.ax[0, 0]
key = ["3337", "1590", "1377"][0]
for img in util.images():
img = util.pad(img, 100, 100)
# angle = PCA_angle(img) / tau * 360
# img = rotate(img, angle=angle)
img = shift(img, 20, 50)
ax.clear()
ax.imshow(img, cmap="gray")
fig.update()
#--- PCA angles
if False: