search

vaibhavsaxena11 / cwvae

Clockwork Variational Autoencoder
http://danijar.com/cwvae
MIT License
45 stars 9 forks source link

Some problems with the accuracy of reproduction #6

Open fengpany opened 1 year ago

fengpany commented 1 year ago

I reproduced the tensorflow version of the CWVAE code, but the accuracy of the reproduction is very different from the paper. Specifically, when testing the minerl_nvigate dataset (where num_examples=225, num_samples=1), the PSNR and SSIM are 0.55 and 17.3, respectively (which I averaged all the frames predicted in the minerl_nvigate test set), but this is much different from the results of the paper of 0.65 and 23.23. Where is the problem?How can I get thesis accuracy？ image

vaibhavsaxena11 commented 1 year ago

Hey, thank you for your interest in this work. Is this a result of training using this repository or your re-implementation?

fengpany commented 1 year ago

Hey, thank you for your interest in this work. Is this a result of training using this repository or your re-implementation?

Yes,this is the result of replicating this repository code.I can't reproduce to the accuracy of the paper.

fengpany commented 1 year ago

Do you have a weight file that achieves the accuracy of the paper, can you send it to me?I want to reproduce to the accuracy of the paper to facilitate my subsequent research.