I have trained a convolutional autoencoder on cifar10 dataset. The reconstruction loss on the test data is quite less (around 0.0225). However, the reconstructed training images do not look like training images (reconstructed images do look similar but with significant loss of information).

Is it a normal thing ? If not, how to overcome this ?

Anybody having practically experienced such a thing ?

I am using Keras library to implement CAE using python.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.