Is there a script for converting the labels into grayscale as required by Caffe?

[Asheesh]

In the Readme, it is mentioned that:

Data

Our example models are trained on the ISPRS Vaihingen dataset. We use the IRRG tiles (8bit format). The ground truth files are color-encoded and should be converted to the numerical labels, e.g. {0,1,2,3,4,5} instead of {[255,255,255],[0,0,255],[0,255,255],[0,255,0],[255,0,255],[255,0,0]}.

This doesn't appear to be happening in any of the notebooks.

[Nicolas Audebert]

Yes, I've been meaning to add those scripts.

In the meantime, you can use those two functions to perform the conversion from RGB-color encoding to grayscale labels :

def convert_to_color(arr_2d):
    """ grayscale labels to RGB-color encoding """
    arr_3d = np.zeros((arr_2d.shape[0], arr_2d.shape[1], 3), dtype=np.uint8)
    palette = {0 : (255, 255, 255), # Impervious surfaces (white)
               1 : (0, 0, 255), # Buildings (dark blue)
               2 : (0, 255, 255), # Low vegetation (light blue)
               3 : (0, 255, 0), # Tree (green)
               4 : (255, 255, 0), # Car (yellow)
               5 : (255, 0, 0), # Clutter (red)
               6 : (0, 0, 0)} # Unclassified (black)

    for c, i in palette.items():
        m = arr_2d == c
        arr_3d[m] = i

    return arr_3d

def convert_from_color(arr_3d):
    """ RGB-color encoding to grayscale labels """
    arr_2d = np.zeros((arr_3d.shape[0], arr_3d.shape[1]), dtype=np.uint8)
    palette = {(0, 0, 0): 0, # Impervious surfaces (white)
                    (0, 0, 255): 1, # Buildings (dark blue)
                    (0, 255, 255): 2, # Low vegetation (light blue)
                    (0, 255, 0): 3, # Tree (green)
                    (255, 255, 0): 4, # Car (yellow)
                    (255, 0, 0): 5, # Clutter (red)
                    (0, 0, 0): 6} # Unclassified (black)

    for c, i in palette.items():
        m = np.all(arr_3d == np.array(c).reshape(1, 1, 3), axis=2)
        arr_2d[m] = i

    return arr_2d

[Nicolas Audebert]

You can now use the convert_gt.py script :

usage: convert_gt.py [-h] [--to-color] [--from-color] [--out OUT]
                     images [images ...]