@@ -6,7 +6,7 @@ This code is the implementation of the [PESTO paper](https://arxiv.org/abs/2309.
...
@@ -6,7 +6,7 @@ This code is the implementation of the [PESTO paper](https://arxiv.org/abs/2309.
that has been accepted at [ISMIR 2023](https://ismir2023.ismir.net/).
that has been accepted at [ISMIR 2023](https://ismir2023.ismir.net/).
**Disclaimer:** This repository contains minimal code and should be used for inference only.
**Disclaimer:** This repository contains minimal code and should be used for inference only.
If you want full implementation details or want to use PESTO for research purposes, take a look at ~~[this repository](https://github.com/aRI0U/pesto-full)~~ (work in progress).
If you want full implementation details or want to use PESTO for research purposes, take a look at ~~[this repository](https://github.com/aRI0U/pesto-full)~~ (coming soon!).
## Installation
## Installation
...
@@ -59,7 +59,8 @@ Alternatively, one can save timesteps, pitch, confidence and activation outputs
...
@@ -59,7 +59,8 @@ Alternatively, one can save timesteps, pitch, confidence and activation outputs
Finally, you can also visualize the pitch predictions by exporting them as a `png` file (you need `matplotlib` to be installed for PNG export). Here is an example:
Finally, you can also visualize the pitch predictions by exporting them as a `png` file (you need `matplotlib` to be installed for PNG export). Here is an example:
Alternatively, one can use basic argmax of weighted average with option `-r`/`--reduction`.
Alternatively, one can use basic argmax of weighted average with option `-r`/`--reduction`.
...
@@ -150,11 +152,11 @@ Note that batched predictions are available only from the Python API and not fro
...
@@ -150,11 +152,11 @@ Note that batched predictions are available only from the Python API and not fro
## Performances
## Performances
On [MIR-1K]() and [MDB-stem-synth](), PESTO outperforms other self-supervised baselines.
On [MIR-1K](https://zenodo.org/record/3532216#.ZG0kWhlBxhE) and [MDB-stem-synth](https://zenodo.org/records/1481172), PESTO outperforms other self-supervised baselines.
Its performances are close to CREPE's, which has 800x more parameters and was trained in a supervised way on a vast
Its performances are close to CREPE's, which has 800x more parameters and was trained in a supervised way on a vast
dataset containing MIR-1K and MDB-stem-synth, among others.
dataset containing MIR-1K and MDB-stem-synth, among others.