Are the released model weights the same as the SVM version mentioned in the paper?

grip-unina / ClipBased-SyntheticImageDetection

Apache License 2.0

12 stars 5 forks source link

Are the released model weights the same as the SVM version mentioned in the paper? #4

Open iamwangyabin opened 1 month ago

iamwangyabin commented 1 month ago

Because I noticed that the released model uses a linear layer instead of an SVM, could you please clarify which version of the model has been released?

davin11 commented 1 month ago

The released version is relative to the SVM. The weights of linear SVM were converted in a pytorch linear layer to simplify the test code.