search

Chris210634 / metric-learning-using-contextual-similarity

17 stars 0 forks source link

why regression loss is not included in the code? #1

Closed sungmin9939 closed 11 months ago

sungmin9939 commented 11 months ago

hi

I'm impressed your wonderful work!

while I am checking your code and paper, what I have found is that regression loss is not applied in the HybridLoss class (the class in 224x224/losses.py file)

Is it just a mistake?

Chris210634 commented 11 months ago

No, the third loss added to the output of the HybridLoss in the following line of train.py https://github.com/Chris210634/metric-learning-using-contextual-similarity/blob/8433dcb67c2205c0e30ec07ed1e5b2fb92da016d/224x224/train.py#L305C21-L305C21

sungmin9939 commented 11 months ago

oh sorry, that was my fault I was confused by difference of return values between the function "contrastive jaccard" in 224x224/losses.py and metric-learning-using.../losses.py