What is the shift operation in the SUNRGBD data set for?

[hyx1999]

depth = np.bitwise_or(np.right_shift(depth,3),np.left_shift(depth,16-3))
I am currently reproducing this paper, so I am studying the author's implementation method. I can't understand this code in sunrgbd_dataset.py. Could someone please give me some help?

[jiunyen-ching]

np.right_shift and np.left_shift are referring to the bitshiting operations. The depth image was bitshifted to provide better visualization. However, these pixel values are not the real depth values that we can directly use for calculation operations. So, when loading the data, bitshifting needs to be reversed via
depth = np.bitwise_or(np.right_shift(depth,3),np.left_shift(depth,16-3))

For example, consider this depth image without any bitshifting modification (Image A)

versus a bitshifted depth image (Image B)

Instantly we would be able to make sense of the shapes and depth differences between the objects in Image B compared to A.

In summary, the two operations involved are
A -> B : depth_b = np.bitwise_or(np.right_shift(depth_a, 16-3), np.left_shift(depth_a,3))
B -> A : depth_a = np.bitwise_or(np.right_shift(depth_b,3), np.left_shift(depth_b,16-3))