search

vadel / AudioUniversalPerturbations

12 stars 6 forks source link

about the end-to-end differentiable #5

Open yveszq opened 3 years ago

yveszq commented 3 years ago

However, there are some modifications in order to ensure that the process is end-to-end differentiable, including the MFCC transformation.

So can you help me with this part of code to ensure that the process is end-to-end differentiable?

Thank you very much！

vadel commented 3 years ago

We followed the strategy proposed by Carlini and Wagner in their implementation for DeepSpeech. You can check their code and all the details in their repository (and in particular in this function).

yveszq commented 3 years ago
Thank you very much for your prompt reply. I will follow their repository as you mentioned. I believe that your helpful guidance will help me a lot with my work. 张强

zq308297543@126.com | 签名由网易邮箱大师定制 On 11/19/2020 19:10，vadelnotifications@github.com wrote：

We followed the strategy proposed by Carlini and Wagner in their implementation for DeepSpeech. You can check their code and all the details in their repository (and in particular in this function).

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.