search

dontLoveBugs / SupervisedDepthPrediction

Pytorch framework for supervised depth prediction
41 stars 15 forks source link

Calculating prob in ordinal regression #6

Open g0401828t opened 3 years ago

g0401828t commented 3 years ago

Hi, thank you for the nice code. It was really nice to refer to. I am a fresh man in deep learning so if the question is dumb, please excuse me.

I am not sure but in the paper, the ord_prob is calculated as below by taking the probability of the second one. image

However, from your code(class OrdinalRegressionLayer), the Order_prob is calculated with the first one. image

Shouldn't 0 be 1 to take the second prob as ord_prob? Or is it that it doesn't matter much to influence the result?

heyuhhh commented 1 year ago

Hi, thank you for the nice code. It was really nice to refer to. I am a fresh man in deep learning so if the question is dumb, please excuse me.

I am not sure but in the paper, the ord_prob is calculated as below by taking the probability of the second one. image

However, from your code(class OrdinalRegressionLayer), the Order_prob is calculated with the first one. image

Shouldn't 0 be 1 to take the second prob as ord_prob? Or is it that it doesn't matter much to influence the result?

I have the same question. Have you solved this problem? @g0401828t