madhavmk / Noise2Noise-audio_denoising_without_clean_training_data

Source code for the paper titled "Speech Denoising without Clean Training Data: a Noise2Noise Approach". Paper accepted at the INTERSPEECH 2021 conference. This paper tackles the problem of the heavy dependence of clean speech data required by deep learning based audio denoising methods by showing that it is possible to train deep speech denoising networks using only noisy speech samples.
MIT License
172 stars 42 forks source link

requirements.txt for Linux #3

Closed HarisNaveed17 closed 3 years ago

HarisNaveed17 commented 3 years ago

Hello, First of all great work!

Can you please add a requirements.txt file for linux as well? The present one does not work directly with the conda create command. This may also be because a lot of packages require channels different from the base channels, if you can specify those channels in the readme that would make it more convenient.

Thank you!

madhavmk commented 3 years ago

Hey,

We use just 2 conda channels: conda-forge and defaults. You are env creation fails because I generated the YAML using Windows platform , but you are transferring across to Linux platform. Solution is to omit build info Convert "cryptography=3.4.7=py38hd7da0ea_0" -> "cryptography=3.4.7" and so on

HarisNaveed17 commented 3 years ago

I already did that but it says the following packages are not available from either channel:

I haven't done development work on Windows so not sure if these packages are platform specific, in that case it makes sense that they're not available on the linux-64 channel. Can you please confirm?

Thanks!

madhavmk commented 3 years ago

No issue. If the library is not on conda, try pip installing them manually one by one. Some libraries are OS specific however. They might not be needed/alternatives need to be found if you are switching OS.