Commit 5191d6d2 authored by cc215's avatar cc215 💬
Browse files

update

parent 36f6a2d5
......@@ -125,13 +125,10 @@ e.g.
- A model trained on UKBB data (SAX slices) with adversarial data augmentation (adversarial noise, adversarial bias field, adversarial morphological deformation, and adversarial affine transformation) is available.
- This model is expected with improved robustness especially for right ventricle segmentation on cross-domain data. See below test results on intra domain test set (UKBB) and _unseen_ cross domain sets ACDC and M\&Ms.
| Testing config <br> (batch_size=1, roi =256) | UKBB test (600) | | | ACDC (100) | | | M\&Ms (150) | | |
| :------------------------------------------: | :------------------------------------------------------------------------------------: | :----: | :----: | :--------: | :----: | :----: | :---------: | :----: | :----: | ------ | --- | --- |
| model: UNet_64 | LV | MYO | RV | LV | MYO | RV | LV | MYO | RV |
| baseline | 0.9383 | 0.8780 | 0.8979 | 0.8940 | 0.8034 | 0.8237 | 0.8862 | 0.7889 | 0.8168 |
| <!-- | Finetune w. random DA | 0.9378 | 0.8768 | 0.8975 | 0.8884 | 0.7981 | 0.8295 | 0.8846 | 0.7893 | 0.8158 |
| <!-- | Finetune w. random DA + adv Bias:<br>UNet*LVSA_Adv_bias*(epochs=20).pth | 0.9326 | 0.8722 | 0.8973 | 0.8809 | 0.7912 | 0.8395 | 0.8794 | 0.7812 | 0.8228 | --> |
| <!-- | Finetune w. random DA + adv Composite DA:<br><br>UNet*LVSA_Adv_Compose*(epochs=20).pth | 0.9360 | 0.8726 | 0.8966 | 0.8984 | 0.7973 | 0.8440 | 0.8873 | 0.7859 | 0.8343 | --> | --> |
| Testing config <br> (batch_size=1, roi =256) | UKBB test (600) | | | ACDC (100) | | | M\&Ms (150) | | |
| :------------------------------------------: | :-------------: | :----: | :----: | :--------: | :----: | :----: | :---------: | :----: | :----: | ---- | --------------------- | ------ | ------ | ------ | ------ | ------ | ------ | ------ | ------ | ------ | --- | ----------------------------------------------------------------------- | ------ | ------ | ------ | ------ | ------ | ------ | ------ | ------ | ------ | --- | -------------------------------------------------------------------------------------- | ------ | ------ | ------ | ------ | ------ | ------ | ------ | ------ | ------ | --- |
| model: UNet_64 | LV | MYO | RV | LV | MYO | RV | LV | MYO | RV |
| baseline | 0.9383 | 0.8780 | 0.8979 | 0.8940 | 0.8034 | 0.8237 | 0.8862 | 0.7889 | 0.8168 | <!-- | Finetune w. random DA | 0.9378 | 0.8768 | 0.8975 | 0.8884 | 0.7981 | 0.8295 | 0.8846 | 0.7893 | 0.8158 | | Finetune w. random DA + adv Bias:<br>UNet*LVSA_Adv_bias*(epochs=20).pth | 0.9326 | 0.8722 | 0.8973 | 0.8809 | 0.7912 | 0.8395 | 0.8794 | 0.7812 | 0.8228 | | Finetune w. random DA + adv Composite DA:<br><br>UNet*LVSA_Adv_Compose*(epochs=20).pth | 0.9360 | 0.8726 | 0.8966 | 0.8984 | 0.7973 | 0.8440 | 0.8873 | 0.7859 | 0.8343 | --> |
| Finetune w. random DA + adv chain:<br><br> (checkpoints/LVSA/Unet_adv_chain.pth) | 0.9360 | 0.8732 | 0.8965 | 0.9060 | 0.8087 | 0.8404 | 0.8929 | 0.7987 | 0.8245 |
......
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment