Late submission results for output of main.py aren't matching reported values

[max-ilse]

Hey,

I just ran main.py and submitted the .csv to kaggle. It tells me that the score is 0.62144.

I was under the impression that the pretrained models can be used to reproduce the score of 0.39975.

Can anybody please verify this finding?

I am aware of the fact that this repo is not claiming that the models can reproduce the results.

Thanks.

[lfz]

the pretrained model should have a score of 0.399, it is verified by many people.

if you trained the model by yourself using the script, the score should also be around 0.41.

please check the intermediate results to see if anything wrong

2018-05-08 0:21 GMT+08:00 max-ilse notifications@github.com:

Hey,

I just ran main.py and submitted the .csv to kaggle. It tells me that the score is 0.62144.

I was under the impression that the pretrained models can be used to reproduce the score of 0.39975.

Can anybody please verify this finding?

I am aware of the fact that this repo is not claiming that the models can reproduce the results.

Thanks.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/lfz/DSB2017/issues/76, or mute the thread https://github.com/notifications/unsubscribe-auth/AIigQ-jEjIhpEwsGW9aIAIZZGfbpjpwgks5twHSlgaJpZM4T1P_I .

-- 廖方舟 清华大学医学院 Liao Fangzhou School of Medicine Tsinghua University Beijing 100084 China

[max-ilse]

I should have read the other issues more carefully, since I was under the impression that a late submission on the gaggle website should lead to a score of 0.399. I know now that for a late submission a score of 0.62144 is correct.

Using the correct labels for the test set I was able to reproduce your score, sorry for not reading carefully.

Best, Max

[anruoxi123]

Hello, I am also on the kaggle submitted as a result, but the result is the final score is 0.81, do you find out why?

Thank you very much for my answer.