lfz
commented
6 years ago because the training overfit easily due to the shortage of the number of labeling
2017-09-13 14:29 GMT+08:00 Shaohuai Shi notifications@github.com:
Dear authors,
May I know if there is any consideration why you use dropout in the test phase of the detector? The code is here: https://github.com/lfz/DSB2017/blob/master/net_ detector.py#L120
Best regards, Shaohuai
— 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/45, or mute the thread https://github.com/notifications/unsubscribe-auth/AIigQ2pKl_8muKnPRQsq08TY61Jm3j8nks5sh3Y-gaJpZM4PVo3J .
-- 廖方舟 清华大学医学院 Liao Fangzhou School of Medicine Tsinghua University Beijing 100084 China
mistake...