Tensorflow - deeplab colormap

Question

I'm messing with Semantic Image Segmentation from google's DeepLab . I want to be able change colors for each semantic ( i.e. person , cat etc) . The method that creates the colormap , with PASCAL benchmark is

def create_pascal_label_colormap():
  """Creates a label colormap used in PASCAL VOC segmentation benchmark.

  Returns:
    A Colormap for visualizing segmentation results.
  """
  colormap = np.zeros((256, 3), dtype=int)
  ind = np.arange(256, dtype=int)

  for shift in reversed(range(8)):
    for channel in range(3):
      colormap[:, channel] |= ((ind >> channel) & 1) << shift
    ind >>= 3

  return colormap

I guess if I change the value of ind with another (instead of 2 to have 3) I get different colors. Also , is there another way to get different colors for the semantics ? I just can't seem to guess how that works , how the colormap is created , using shifting as we see in the code . I'm linking also the full code I'm working on from DeepLab,on google colab : https://colab.research.google.com/drive/1a3TnfeEjVMg7N1Dz5d_UA8GN_iKHkG_l#scrollTo=na9DyxVl4_Ul

Answer 1

If you have a fixed number of classes you could also just hard-code the colors you'd like to have, something like

def create_pascal_label_colormap(): 
 return np.asarray([ 
    [0, 0, 0],
    [0, 192, 0],
    ])