search

med-air / DLTTA

[IEEE TMI'22] DLTTA: Dynamic Learning Rate for Test-time Adaptation on Cross-domain Medical Images
16 stars 2 forks source link

about results #1

Closed SakurajimaMaiii closed 2 years ago

SakurajimaMaiii commented 2 years ago

我想问下Table 3 里面 baseline (w/o adaptation) 的结果， Dataset D和 Datset F 是写反了吗 我按照你们的设置和代码train了一下 结果：

Dataset B          C        D        E        F      avg
Dice--- 79.68 73.83 85.01 62.28 67.16  73.59

比论文里高一点点，但是D和F貌似差的有点多？ 想请教一下您的看法

HongzhengYang commented 2 years ago

Hi, thanks for your interest. We have checked the results of Site D and Site F. The processed data of all six sites are uploaded to the google drive (https://drive.google.com/drive/folders/1yol4sUYh7EtQJzfiPn7mNggypBus1IKM?usp=sharing). You can use them to train the model. The results of the two sites are not reversed. Please also note that we only evaluate the prostate center region of each case at test time.

SakurajimaMaiii commented 2 years ago

Got it. thanks your reply.