hsiangyuzhao
commented
6 months ago Sorry for the late reply. Please refer to #issue5 for more details about reproductivity.
Similar hardware and running environments are required to obtain a good reproduction of our results.
Sorry to bother you again. There are some differences between my experimental results and yours. I want to know that is the experimental result of the following figure in the article obtained through the experimental scenario of 'Training with different label ratio '?
Following your experimental instructions, I obtained the following experimental results of the ' Training with different label ratio ' experimental scenario, which seems to be different from yours.
' Training with different label ratio ' experimental scenario： ratio=0.1 bg_dice: 0.9895 ± 0.0075; la_dice: 0.8736 ± 0.0516; bg_hd95: 2.0215 ± 1.2702; la_hd95: 11.6413 ± 11.1788; bg_asd: 0.3815 ± 0.1997; la_asd: 2.7246 ± 1.6591;
ratio=0.2 bg_dice: 0.9918 ± 0.0049; la_dice: 0.9007 ± 0.0313; bg_hd95: 1.668 ± 0.8762; la_hd95: 6.5163 ± 3.4624; bg_asd: 0.287 ± 0.1349; la_asd: 1.9518 ± 0.8151;