Train result is not consistent with the original paper

dansuh17 / segan-pytorch

SEGAN pytorch implementation https://arxiv.org/abs/1703.09452

GNU General Public License v3.0

106 stars 32 forks source link

Train result is not consistent with the original paper #16

Closed xinkez closed 4 years ago

xinkez commented 6 years ago

Hi, Thanks for your codes. I use your codes to train the segan network. When the result came out, it is not consistent with the original paper. The noisy signal is denoised less. Have you compared your results with the original paper before? Thanks in advance.

dansuh17 commented 5 years ago

I have compared them, and it worked fine. However, it has been a while since I did that. Somehow, the codes in this repository are somewhat different from the code I used for testing. There are plans to refine this project soon (pytorch 1.0 is coming!). I will generate and post a link to samples generated by then :)