search

daQuincy / VAE-GAN-Autoencoding-Beyond-Pixels-Using-a-Similarity-Metric

Simple Tensorflow implementation of the paper Autoencoding Beyond Pixels Using a Similarity Metric
MIT License
8 stars 0 forks source link

Wrong output for GAN #3

Open At-Walid opened 2 years ago

At-Walid commented 2 years ago

Hello,

I tried your notebook with another dataset, I'm getting the image below when running GAN and GAN_VAE. Do you have any idea of what can be the problem?

Thank you

image
daQuincy commented 2 years ago

Hi,

May I know what dataset are you using? This is completely normal as the training for GAN and VAE is always unstable. When posterior collapse happens the resulting image will be what you just showed me here.

On Wed, 15 Dec 2021 at 21:19, At-Walid @.***> wrote:

Hello,

I tried your notebook with another dataset, I'm getting the image below when running GAN and GAN_VAE. Do you have any idea of what can be the problem?

Thank you [image: image] https://user-images.githubusercontent.com/63915243/146193483-1fd435bc-945c-48e9-9bdf-f66baa4c5e26.png

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/daQuincy/VAE-GAN-Autoencoding-Beyond-Pixels-Using-a-Similarity-Metric/issues/3, or unsubscribe https://github.com/notifications/unsubscribe-auth/AEZJN54IZH7ZX3YOAMLWBXLURCIWXANCNFSM5KDTX7WA .

At-Walid commented 2 years ago

Hello,

thank you for your quick answer,

I used data issued from a simulator, here is an example:

image

Is there any modification of the code I need to do so that it works?

Thank you,

daQuincy commented 2 years ago

Short answer is there is no simple solution to this problem. The training of VAE and GAN is known to be unstable. This requires a lot of understanding in the formulation of VAE/GAN, research effort and maybe hyperparameter tuning and understanding your dataset.

I apologise for not being able to be more helpful as this topic is vast and complex.

On Thu, 16 Dec 2021 at 18:45, At-Walid @.***> wrote:

Hello,

thank you for your quick answer,

I used data issued from a simulator, here is an example: [image: image] https://user-images.githubusercontent.com/63915243/146357136-9fb1810c-4653-4f0d-a63d-a1ce530e184f.png

Is there any modification of the code I need to do so that it works?

Thank you,

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/daQuincy/VAE-GAN-Autoencoding-Beyond-Pixels-Using-a-Similarity-Metric/issues/3#issuecomment-995652982, or unsubscribe https://github.com/notifications/unsubscribe-auth/AEZJN572OOPZ2ARH3F5Z4RLURG7M5ANCNFSM5KDTX7WA .