Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
CREPE: A Convolutional REpresentation for Pitch Estimation -- pre-trained model (ICASSP 2018)
| Date | Stars |
|---|---|
| 2026-07-24 | 1408 |
| 2026-07-25 | 1408 |
| 2026-07-28 | 1408 |
| 2026-07-30 | 1408 |
| 2026-08-06 | 1408 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
CREPE Pitch Tracker
===================
[](https://pypi.python.org/pypi/crepe)
[](https://opensource.org/licenses/MIT)
[](https://travis-ci.org/marl/crepe)
[](https://pepy.tech/project/crepe)
[]()
<!--[](https://coveralls.io/github/marl/crepe?branch=master)
[](http://crepe.readthedocs.io/en/latest/?badge=latest)-->
CREPE is a monophonic pitch tracker based on a deep convolutional neural network operating directly on the time-domain waveform input. CREPE is state-of-the-art (as of 2018), outperfoming popular pitch trackers such as pYIN and SWIPE:
<p align="center"><img src="https://user-images.githubusercontent.com/3009670/36563051-ee6a69a0-17e6-11e8-8d7b-9a37d16ee7ad.png" width="500"></p>
Further details are provided in the following paper:
> [CREPE: A Convolutional Representation for Pitch Estimation](https://arxiv.org/abs/1802.06182)<br>
> Jong Wook Kim, Justin Salamon, Peter Li, Juan Pablo Bello.<br>
> Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), 2018.
We kindly request that academic publications making use of CREPE cite the aforementioned paper.
## Installing CREPE
CREPE is hosted on PyPI. To install, run the following command in your Python environment:
```bash
$ pip install --upgrade tensorflow # if you don't already have tensorflow >= 2.0.0
$ pip install crepe
```
To install the latest version from source clone the repository and from the top-level `crepe` folder call:
```bash
$ python setup.py install
```
## Using CREPE
### Using CREPE from the command line
This package includes a command line utility `crepe` and a pre-trained version of the CREPE model for easy use. To estimate the pitch of `audio_file.wav`, run:
```bash
$ crepe audio_file.wav
```
or
```bash
$ python -m crepe audio_file.wav
```
The resulting `audio_file.f0.csv` contains 3 columns: the first with timestamps (a 10 ms hop size is used by default), the second contains the predicted fundamental frequency in Hz, and the third contains the voicing confidence, i.e. the confidence in the presence of a pitch:
time,frequency,confidence
0.00,185.616,0.907112
0.01,186.764,0.844488
0.02,188.356,0.798015
0.03,190.610,0.746729
0.04,192.952,0.771268
0.05,195.191,0.859440
0.06,196.541,0.864447
0.07,197.809,0.827441
0.08,199.678,0.775208
...
#### Timestamps
CREPE uses 10-millisecond time steps by default, which can be adjusted using
the `--step-size` option, which takes the size of the time step in millisecond.
For example, `--step-size 50` will calculate pitch for every 50 milliseconds.
Following the convention adopted by popular audio processing libraries such as
[Essentia](http://essentia.upf.edu/) and [Librosa](https://librosa.github.io/librosa/),
from v0.0.5 onwards CREPE will pad the input signal such that the first frame
is zero-centered (the center of the frame corresponds to time 0) and generally
all frames are centered around their corresponding timestamp, i.e. frame
`D[:, t]` is centered at `audio[t * hop_length]`. This behavior can be changed
by specifying the optional `--no-centering` flag, in which case the first frame
will *start* at time zero and generally frame `D[:, t]` will *begin* at
`audio[t * hop_length]`. Sticking to the default behavior (centered frames) is
strongly recommended to avoid misalignment with features and annotations produced
by other common audio processing tools.
#### Model Capacity
CREPE uses the model size that was reported in the paper by dExcerpt of 8,291 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:766c0234b8110a3c, topic:tensorflow
