Closed melody-rain closed 5 years ago
Many researchers have reproduced the work. You may read the previous issues to solve the OOM. If you have results, please reply here so that we can help ours. Thanks!
@wentaozhu
I forgot to combine the detection csvs into one. After combining, I got similar results to your paper.
Thanks.
Do you got get 1-fold performance by concatenating the result csvfiles which are in the 150 val?Or you concatenated the 10 csvfiles which are the results of all 10-fold in every val150 file to get the fina performance?
later one
Could anyone reproduce the author's result?
I trained the model according to this repo with training data of luna subset 0~8 and test the model regarding luna subset 9 as the testing data.
By changing below to avoid GPU OOM:
my testing result is similar to this:
Even though I use the provided model to test,
I just got poor sensitivity which is less than 0.1.
So have anyone reproduced the paper? Thanks.