Evaluation on BCSS dataset about ads_unet

[ynulonger]

Sorry for the later response. The following command can be used to run testing:
python -u predict_ada_hist.py -w checkpoints/BCSS/ds_True_Ada_scse/0_0.001_Ada_scse -o scse -l 1234_pred -v soft -d BCSS

• -w path to the checkpoints, where you save the trained models.
• -o feature re-calibration in skip connection, option: [scse, None].
• -l specify the output from the ADS_UNet. option:

  1234_pred, Collecting softmax prediction from each base $UNet^d$
  1, Only the output of the UNet$^1$ is returned
  ...
  4, Only the output of the UNet$^4$ is returned

• -v voting strategy, option:

  soft, $\alpha$ voting
  mean, average voting

• -d dataset, option: [BCSS, CRAG, Kumar]
  Still working on cleaning the code and will update the README later, apologize for bring you confusion.

Best,

[guascy666]

Hi,thanks for your code,i want to konw that how to generate the whole image from the list of the cropped patch image