def runTexture(img_list):
""" This function administrates the extraction of a video texture from the
given frames.
"""
video_volume = assignment11.videoVolume(img_list)
ssd_diff = assignment11.sumSquaredDifferences(video_volume)
transition_diff = assignment11.transitionDifference(ssd_diff)
alpha = 1.5*10**6
idxs = assignment11.findBiggestLoop(transition_diff, alpha)
diff3 = np.zeros(transition_diff.shape, float)
for i in range(transition_diff.shape[0]):
for j in range(transition_diff.shape[1]):
diff3[i,j] = alpha*(i-j) - transition_diff[i,j]
return vizDifference(ssd_diff), \
vizDifference(transition_diff), \
vizDifference(diff3), \
assignment11.synthesizeLoop(video_volume, idxs[0]+2, idxs[1]+2)
def test_transitionDifference():
ssd1 = np.zeros((9,9), dtype = float)
ssd1[4,4] = 1
ssd2 = np.eye(5, dtype = float)
out1 = np.array([[0.0625, 0. , 0. , 0. , 0. ],
[0. , 0.25, 0. , 0. , 0. ],
[0. , 0. , 0.375, 0. , 0. ],
[0. , 0. , 0. , 0.25, 0. ],
[0. , 0. , 0. , 0. , 0.0625]], dtype = float)
out2 = np.array([[1.]], dtype = float)
for ssd, true_out in zip((ssd1, ssd2), (out1, out2)):
print "Input:\n{}\n".format(ssd)
usr_out = assignment11.transitionDifference(ssd)
if type(usr_out) != type(true_out):
print ("Error: transitionDifference output has type {}. " +
"Expected type is {}.").format(type(usr_out), type(true_out))
return False
if usr_out.shape != true_out.shape:
print ("Error: transitionDifference output has shape {}. " +
"Expected shape is {}.").format(usr_out.shape,
true_out.shape)
return False
if usr_out.dtype != true_out.dtype:
print ("Error: transitionDifference output has dtype {}. " +
"Expected dtype is {}.").format(usr_out.dtype,
true_out.dtype)
return False
if not np.all(np.abs(usr_out - true_out) < 0.10):
print ("Error: transitionDifference output has value:\n{}\n" +
"Expected value:\n{}").format(usr_out, true_out)
return False
print "Passed current input."
print "transitionDifference passed."
return True
def runTexture(img_list, alpha):
""" This function administrates the extraction of a video texture from the
given frames, and generates the three viewable difference matrices.
"""
video_volume = a11.videoVolume(img_list)
ssd_diff = a11.computeSimilarityMetric(video_volume)
transition_diff = a11.transitionDifference(ssd_diff)
print "Alpha is {}".format(alpha)
idxs = a11.findBiggestLoop(transition_diff, alpha)
diff3 = np.zeros(transition_diff.shape, float)
for i in range(transition_diff.shape[0]):
for j in range(transition_diff.shape[1]):
diff3[i, j] = alpha * (i - j) - transition_diff[i, j]
return (vizDifference(ssd_diff),
vizDifference(transition_diff),
vizDifference(diff3),
a11.synthesizeLoop(video_volume, idxs[0] + 2, idxs[1] + 2))
