weights balancing for multilabels data

[tanveer6715]

Hi,
Good work. This code is for binary segmentation. If we have 4 or 5 classes in dataset then how can we calculate the weights of each class?

[yhlleo]

Hi @tanveer6715 , thanks. The calculation for multiple classes is also similar:

Given [w_1, w_2, ..., w_k], where w_i refers to the percentage of the frequency of class i, we can have the median w_med = median([w_1, w_2, ..., w_k]), then the can balance the weights in this manner: w'_i = w_med/w_i.