u-net-brain-tumor
u-net-brain-tumor copied to clipboard
DICE
I got dice equal to one every epoch? How to handle that?
check the buttom of the readme
I meet the same question. When I begin the code "python train.py --task=necrotic", at the beginning, the dice is between 0.7 and 0.8, however, after about 30 epoches, the dice becomes 1 all the time. My data_size = "small". I sincerely hope that someone can give me some constructive suggestions! @zsdonghao @ywangc
@cb632482135qq I had this same issue, did a few adjustments to the training, but what really solved the problem was lowering the learning rate, in my case I added one more zero (lowered it by 10, perhaps too much). The learning rate is later updated, so this is only a starting learning rate, therefore it should be alright.
Can u please provide me the outputs u came across. Specially, images showing the tumorous parts represented in different respwctive colors.
Can u please provide me the outputs u came across. Specially, images showing the tumorous parts represented in different respwctive colors.
I think it is in opencv