Issues on function: DS3_Dempster()

[xuanlongORZ]

Thank you for sharing the code, it is really cool.
However, I found that it might have a mistake in the function DS3_Dempster().
In your paper: Evidential fully convolutional network for semantic segmentation, the way that the combination of the masses given as in Eq(6a) and Eq (6b) is different from the codes in this function. For example on line 125:
m1=inputs[:,:,:,0,:] should be m1=inputs[:,:,:,0,:-1] and so on.
Also on line 136:
omega1=tf.expand_dims(combine2_3[:,:,:,-1], -1) should be omega1=omega1 * omega2 as shown in Eq (6b).

Please let me know if I misunderstood anything, thanks a lot.