OpenGAN
OpenGAN copied to clipboard
aimerykong
hi, i have a question about the code
hi, i see the test.ipynb ,the cls model is no use in the ipynb; and i do not know if it is just train GAN model first(use outlier data and close set data ),and extract features from pretrained model and put in the D model to output value in the test...
Hi, thanks for your interest. You are right! The "cls" is the K-way classifier for the closed-set classification. OpenGAN repurposes the off-the-shelf features and learns a discriminator for open-set detection. Therefore, cls does not play a role here.
On Wed, Nov 3, 2021 at 8:07 AM 1senlin @.***> wrote:
hi, i see the test.ipynb ,the cls model is no use in the ipynb; and i do not know if it is just train GAN model first(use outlier data and close set data ),and extract features from pretrained model and put in the D model to output value in the test...
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/aimerykong/OpenGAN/issues/5, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABRJSJCFHTKEFZBAVMLU66TUKEQYVANCNFSM5HIYBGTA . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.
Hi, thanks for your interest. You are right! The "cls" is the K-way classifier for the closed-set classification. OpenGAN repurposes the off-the-shelf features and learns a discriminator for open-set detection. Therefore, cls does not play a role here. … On Wed, Nov 3, 2021 at 8:07 AM 1senlin @.***> wrote: hi, i see the test.ipynb ,the cls model is no use in the ipynb; and i do not know if it is just train GAN model first(use outlier data and close set data ),and extract features from pretrained model and put in the D model to output value in the test... — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#5>, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABRJSJCFHTKEFZBAVMLU66TUKEQYVANCNFSM5HIYBGTA . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.
thanku, i have another question, in the paper ,i see the loss should have open training samples loss, but i do not see this in the code ,is it necessary add this if our training data contains open training samples?
Besides, i try the training demo more than once, i find the same epochs models(for example 60 epoch) have different results (AUC) when i test with the same dataset...
Yes, adding open training samples is much better. Please refer to the second last cell for how to do so ( https://github.com/aimerykong/OpenGAN/blob/main/demo_OpenSetSegmentation_training.ipynb )
When you say "the same epochs models have different results", I guess you are referring to the cross-dataset GAN training? Does random initialization affect performance? What are the AUC's you got from your two models?
On Wed, Nov 3, 2021 at 9:34 AM 1senlin @.***> wrote:
Besides, i try the training demo more than once, i find the same epochs models(for example 60 epoch) have different results (AUC) when i test with the same dataset...
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/aimerykong/OpenGAN/issues/5#issuecomment-959089891, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABRJSJBI2WJ6MPFSHNAY6JTUKE26BANCNFSM5HIYBGTA . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.