search

bigpon / vcc20_baseline_cyclevae

Voice Conversion Challenge 2020 CycleVAE baseline system
MIT License
133 stars 17 forks source link

Which data set should I use #5

Closed user-ZJ closed 3 years ago

user-ZJ commented 3 years ago

Hi I want to rerun your code, but I have some questions.

At baseline readme.md:

Then, move to the experiment directory, and copy the waveform data of VCC 2020 to wav_24kHz folder.
$ cd ../egs/cyclevae
$ cp -r <path_to_vcc2020_data>/* wav_24kHz

At the main readme.md,have there dataset:

2020/4/17 upload the missed conversion pair of SEF2-TEM1 of reference_v.10.

2020/3/18 release the generated samples of reference_v.10.

2020/3/11 release the first version repo and the generated samples of development set (dv50_vcc2020_24kHz).

My question is Which data set should I use. thanks!

bigpon commented 3 years ago

Hi ZJ, We trained our models based on the Voice Conversion Challenge 2020 corpus. However, the organizers haven't publically released the corpus now (They have released VCC2018 and VCC2016). Maybe you can contact the organizers to access the corpus. The website of VCC2020 is the following, http://www.vc-challenge.org/ or you may try VCC2018 corpus first. https://datashare.is.ed.ac.uk/handle/10283/3061

user-ZJ commented 3 years ago

@bigpon Could you share the data you got, thanks! I just want to reproduce "Baseline System of Voice Conversion Challenge 2020 with Cyclic Variational Autoencoder and ParallelWaveGAN" paper

bigpon commented 3 years ago

Hi ZJ, I'm sorry that we should follow the schedule of the VCC20 organizers. We cannot release the VCC20 corpus by ourselves. I suggest that contacting the organizers for requiring the corpus.

user-ZJ commented 3 years ago

thanks