Image Size and Loss function
53RT opened this issue · 2 comments
53RT commented
Hi,
thanks for publishing your code and congratulation for your great work.
There are two things I would like to know but couldn't find them in the breast density paper nor in the Bi-Rads paper.
- What is the typical image size in your dataset before cropping?
- Which loss function have you used in your training?
kjgeras commented
- I don't remember the exact number, it would be roughly 20% bigger on each side.
- cross_entropy, i.e. simply -log y_correct.
53RT commented
thank you for the instant answer