Fréchet Audio Distance: A Reference-Free Metric for Evaluating Music Enhancement Algorithms

By Kevin Kilgour, Mauricio Zuluaga, Dominik Roblek, Matthew Sharifi. 2019

342 citations recorded by the source corpus.

We propose the Fréchet Audio Distance (FAD), a novel, reference-free evaluation metric for music enhancement algorithms. We demonstrate how typical evaluation metrics for speech enhancement and blind source separation can fail to accurately measure the perceived effect of a wide variety of distortions. As an alternative, we propose adapting the Fréchet Inception Distance (FID) metric used to evaluate generative image models to the audio domain. FAD is validated using a wide variety of artificial distortions and is compared to the signal based metrics signal to distortion ratio (SDR), cosine distance, and magnitude L2 distance. We show that, with a correlation coefficient of 0.52, FAD correlates more closely with human perception than either SDR, cosine distance or magnitude L2 distance, with correlation coefficients of 0.39, −0.15 and −0.01 respectively.

Subfields: Computer Vision and Pattern Recognition, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Signal Processing

Citations in the corpus

Audiobox: Unified Audio Generation with Natural Language Prompts (2023, methodology): “Finally, comparing the last two rows of Table 7 suggests reranking with CLAP model is an effective approach to improving the overall performance in both the audio quality (FAD: 0 .”
AudioGen: Textually Guided Audio Generation (2022, methodology): “While showing inferior FAD and KL results when compared to the single-stream model, the multi-stream setting offers a trade-off between inference quality and speed.”
MusicLM: Generating Music From Text (2023, background): “The Fr´echet Audio Distance (Kilgour et al., 2019) is a reference-free audio quality metric, which correlates well with human perception.”
AudioLDM: Text-to-Audio Generation with Latent Diffusion Models (2023, methodology): “In this case, the improvement in the adherence to detailed language description may become misleading information to the classifier in FAD.”
Musika! Fast Infinite Waveform Music Generation (2022, positive): “We also use the Frechét Audio Distance [43] (FAD) metric to quantitatively evaluate the quality of the generated piano samples in Table 2.”
JEN-1: Text-Guided Universal Music Generation with Omnidirectional Diffusion Models (2023, methodology): “The objective evaluation includes three metrics: Fréchet Audio Distance (FAD) (Kilgour et al., 2019), Kullback-Leibler Divergence (KL) (Van Erven & Harremos, 2014), and CLAP score (CLAP) (Elizalde et al.”
Pros and Cons of GAN Evaluation Measures: New Developments (2021, background): “In addition to above, Fréchet Audio Distance (FAD) (Roblek et al., 2019) and Fréchet ChemNet Distance (FCD) (Preuer et al., 2018) have been suggested for evaluation of music enhancement and molecule generation algorithms.”