Leesoon1984
Results
3
issues of
Leesoon1984
I have trained the model for several times, while the results are poorer than the paper's.
While Fine-tuning the distilled model, the performance on the VisDA-C dataset drops, the results are as follows: ``` seed=2019: 74.98 --> 74.80 seed=2020: 76.17 --> 75.17 seed=2021: 79.8 --> 78.6....
For MR-to-CT adaptation, the model is trained on MR and tested on fake MR generated by [CycleGAN](https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix). When [CycleGAN](https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix) is implemented, which image is used, the original [MMWHS Challenge](https://paperswithcode.com/dataset/mm-whs-2017) or...