search

Yanfeng-Zhou / XNet

[ICCV2023] XNet: Wavelet-Based Low and High Frequency Merging Networks for Semi- and Supervised Semantic Segmentation of Biomedical Images
MIT License
184 stars 8 forks source link

About test threshold #17

Open QuintinDong opened 7 months ago

QuintinDong commented 7 months ago

Thanks to the authors for their work! May I ask how much you set the threshold in parser.add_argument('--threshold', default=None) during testing to achieve the best result? I changed None to 0.7 finally tested Dice as 0.7732 on LA dataset and it didn't work very well. What is the threshold setting for your experimental testing procedure?

Yanfeng-Zhou commented 6 months ago

We have discussed this issue in #6. Only binary classification problems have this parameter, and this parameter is printed out during the training process. If you do not have a record, you can set threshold=0.5