search

google-ai-edge / mediapipe

Cross-platform, customizable ML solutions for live and streaming media.
https://ai.google.dev/edge/mediapipe
Apache License 2.0
26.97k stars 5.1k forks source link

Integrating DTLN's Noise Reduction #3355

Closed seanchann closed 2 years ago

seanchann commented 2 years ago

I want to use mediapipe to run denoising of real-time audio (I already have a model trained with tesonrflow), there is no video input, just audio. I didn't find a relevant use case, but is this theoretically possible?

the DTLN project: https://github.com/breizhn/DTLN

sureshdagooglecom commented 2 years ago

Hi @seanchann , Could you provide more information w.r.t above issue.

seanchann commented 2 years ago

There is already a trained tflite model file, then mediapipe read a wav file and uses the trained model (the purpose is to reduce noise), and outputs a denoised wav file

seanchann commented 2 years ago

I mean is it possible to use mediapipe to complete this function（input a wav file for denoise and output new wav file）, because no similar solution was found

sureshdagooglecom commented 2 years ago

Hi @seanchann , MediaPipe is mainly focused on computer vision problems.
The OSS mediapipe will not contain an example or solution for audio in near term.