crlandsc/torch-log-wmse
logWMSE, an audio quality metric & loss function with support for digital silence target. Useful for training and evaluating audio source separation systems.
This tool helps audio engineers and researchers evaluate and improve audio source separation and denoising models. It takes unprocessed, processed, and target audio files as input, and outputs a quality score that accurately reflects human perception, especially for segments containing digital silence. It's designed for anyone working on machine learning models that process audio.
Available on PyPI.
Use this if you are training or evaluating audio processing models, particularly for source separation or denoising, and need a robust quality metric that handles digital silence and aligns with human hearing.
Not ideal if you need a metric that is invariant to arbitrary scaling or polarity inversion, or if you require a full model of human auditory perception like auditory masking.
Stars
45
Forks
1
Language
Python
License
Apache-2.0
Category
Last pushed
Jan 29, 2026
Commits (30d)
0
Dependencies
3
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/voice-ai/crlandsc/torch-log-wmse"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
descriptinc/descript-audio-codec
State-of-the-art audio codec with 90x compression factor. Supports 44.1kHz, 24kHz, and 16kHz...
drethage/speech-denoising-wavenet
A neural network for end-to-end speech denoising
YuanGongND/ast
Code for the Interspeech 2021 paper "AST: Audio Spectrogram Transformer".
iver56/torch-audiomentations
Fast audio data augmentation in PyTorch. Inspired by audiomentations. Useful for deep learning.
lmnt-com/wavegrad
A fast, high-quality neural vocoder.