make_video_from_eigens.py

# Get the eigenvectors
F, T, H, W = activations.shape
points = activations.reshape([F, T*H*W]).transpose()
A = construct_affinity_mat(points, sigmas=[6.0])
eigvecs = wncuts(A, num_eigenvectors=100)
eigvecs_reshaped = eigvecs.reshape([T, H, W, -1])

# frame_shape is [h, w, c]
frame_shape = frames[0].shape[:2]
target_vec = eigvecs_reshaped[0, 0, 0]
l2_diff = np.linalg.norm(eigvecs_reshaped - target_vec, axis=3)

hist = np.histogram(l2_diff, bins=25)
for count, interval in zip(hist[0], hist[1]):
  print('interval ', interval, count)

l2_diff = np.exp(.01 * -l2_diff**2)
histexp = np.histogram(l2_diff, bins=25)
for count, interval in zip(histexp[0], histexp[1]):
  print('interval ', interval, count)

# nearest neighbor from skimage.transform, colors from matplotlib
resized_img = resize(l2_diff, [len(frames), frame_shape[0], frame_shape[1]], order=0)
norm = colors.Normalize(resized_img.min(), resized_img.max())
cmap = plt.cm.RdBu

# get the image w/o the alpha channel.
rgb_img = np.delete(cmap(norm(resized_img)), 3, 3)

for num, (frame, resized) in enumerate(zip(frames, rgb_img)):
  image = Image.fromarray(frame)
  image = image.convert('HSV')
  frame_hsv = np.array(image)

  image = Image.fromarray((resized.squeeze() * 255).astype(np.uint8))
  image = image.convert('HSV')
  resized_hsv = np.array(image)
  
  luma = frame_hsv[:, :, 2:]
  chroma = resized_hsv[:, :, 1:2]

  newimg = np.concatenate([resized_hsv[:, :, 0:1], chroma, luma], 2)
  Image.fromarray(newimg, "HSV").convert(mode="RGB").save('combined.%04d.png' % num)

clips = [ImageClip(os.path.join('combined.%04d.png' % i)).set_duration(1./8) 
         for i in range(len(frames))]
concat_clip = concatenate_videoclips(clips, method="compose")
concat_clip.write_videofile("im-a-videofile.mp4", fps=8)