Efficient neural networks for analog audio effect modeling

Overview

micro-TCN

Efficient neural networks for audio effect modeling.

| Paper | Demo | Plugin |

Setup

Install the requirements.

python3 -m venv env/
source env/bin/activate
pip install -r requirements.txt

Then install auraloss.

pip install git+https://github.com/csteinmetz1/auraloss

Pre-trained models

You can download the pre-trained models here. Then unzip as below.

mkdir lightning_logs
mv models.zip lightning_logs/
cd lightning_logs/
unzip models.zip 

Use the compy.py script in order to process audio files. Below is an example of how to run the TCN-300-C pre-trained model on GPU. This will process all the files in the audio/ directory with the limit mode engaged and a peak reduction of 42.

python comp.py -i audio/ --limit 1 --peak_red 42 --gpu

If you want to hear the output of a different model, you can pass the --model_id flag. To view the available pre-trained models (once you have downloaded them) run the following.

python comp.py --list_models

Found 13 models in ./lightning_logs/bulk
1-uTCN-300__causal__4-10-13__fraction-0.01-bs32
10-LSTM-32__1-32__fraction-1.0-bs32
11-uTCN-300__causal__3-60-5__fraction-1.0-bs32
13-uTCN-300__noncausal__30-2-15__fraction-1.0-bs32
14-uTCN-324-16__noncausal__10-2-15__fraction-1.0-bs32
2-uTCN-100__causal__4-10-5__fraction-1.0-bs32
3-uTCN-300__causal__4-10-13__fraction-1.0-bs32
4-uTCN-1000__causal__5-10-5__fraction-1.0-bs32
5-uTCN-100__noncausal__4-10-5__fraction-1.0-bs32
6-uTCN-300__noncausal__4-10-13__fraction-1.0-bs32
7-uTCN-1000__noncausal__5-10-5__fraction-1.0-bs32
8-TCN-300__noncausal__10-2-15__fraction-1.0-bs32
9-uTCN-300__causal__4-10-13__fraction-0.1-bs32

We also provide versions of the pre-trained models that have been converted to TorchScript for use in C++ here.

Evaluation

You will first need to download the SignalTrain dataset (~20GB) as well as the pre-trained models above. With this, you can then run the same evaluation pipeline used for reporting the metrics in the paper. If you would like to do this on GPU, perform the following command.

python test.py \
--root_dir /path/to/SignalTrain_LA2A_Dataset_1.1 \
--half \
--preload \
--eval_subset test \
--save_dir test_audio \

In this case, not only will the metrics be printed to terminal, we will also save out all of the processed audio from the test set to disk in the test_audio/ directory. If you would like to run the tests across the entire dataset you can specific a different string after the --eval_subset flag, as either train, val, or full.

Training

If would like to re-train the models in the paper, you can run the training script which will train all the models one by one.

python train.py \ 
--root_dir /path/to/SignalTrain_LA2A_Dataset_1.1 \
--precision 16 \
--preload \
--gpus 1 \

Plugin

We provide plugin builds (AV/VST3) for macOS. You can also build the plugin for your platform. This will require the traced models, which you can download here. First, you will need download and extract libtorch. Check the PyTorch site to find the correct version.

wget https://download.pytorch.org/libtorch/cpu/libtorch-macos-1.7.1.zip
unzip libtorch-macos-1.7.1.zip

Now move this into the realtime/ directory .

mv libtorch realtime/

We provide a ncomp.jucer file and a CMakeLists.txt that was created using FRUT. You will likely need to compile and run FRUT on this .jucer file in order to create a valid CMakeLists.txt. To do so, follow the instructions on compiling FRUT. Then convert the .jucer file. You will have to update the paths here to reflect the location of FRUT.

cd realtime/plugin/
../../FRUT/prefix/FRUT/bin/Jucer2CMake reprojucer ncomp.jucer ../../FRUT/prefix/FRUT/cmake/Reprojucer.cmake

Now you can finally build the plugin using CMake with the build.sh script. BUT, you will have to first update the path to libtorch in the build.sh script.

rm -rf build
mkdir build
cd build
cmake .. -G Xcode -DCMAKE_PREFIX_PATH=/absolute/path/to/libtorch ..
cmake --build .

Citation

If you use any of this code in your work, please consider citing us.

    @article{steinmetz2021efficient,
            title={Efficient Neural Networks for Real-time Analog Audio Effect Modeling},
            author={Steinmetz, Christian J. and Reiss, Joshua D.},
            journal={arXiv:2102.06200},
            year={2021}}
Owner
Christian Steinmetz
Building tools for musicians and audio engineers (often with machine learning). PhD Student at Queen Mary University of London.
Christian Steinmetz
The official code for PRIMER: Pyramid-based Masked Sentence Pre-training for Multi-document Summarization

PRIMER The official code for PRIMER: Pyramid-based Masked Sentence Pre-training for Multi-document Summarization. PRIMER is a pre-trained model for mu

AI2 114 Jan 06, 2023
Buffon’s needle: one of the oldest problems in geometric probability

Buffon-s-Needle Buffon’s needle is one of the oldest problems in geometric proba

3 Feb 18, 2022
A PyTorch implementation of "Cluster-GCN: An Efficient Algorithm for Training Deep and Large Graph Convolutional Networks" (KDD 2019).

ClusterGCN ⠀⠀ A PyTorch implementation of "Cluster-GCN: An Efficient Algorithm for Training Deep and Large Graph Convolutional Networks" (KDD 2019). A

Benedek Rozemberczki 697 Dec 27, 2022
Blind Image Super-resolution with Elaborate Degradation Modeling on Noise and Kernel

Blind Image Super-resolution with Elaborate Degradation Modeling on Noise and Kernel This repository is the official PyTorch implementation of BSRDM w

Zongsheng Yue 69 Jan 05, 2023
Practical Blind Denoising via Swin-Conv-UNet and Data Synthesis

Practical Blind Denoising via Swin-Conv-UNet and Data Synthesis [Paper] [Online Demo] The following results are obtained by our SCUNet with purely syn

Kai Zhang 312 Jan 07, 2023
Improving Factual Completeness and Consistency of Image-to-text Radiology Report Generation

Improving Factual Completeness and Consistency of Image-to-text Radiology Report Generation The reference code of Improving Factual Completeness and C

46 Dec 15, 2022
A programming language written with python

Kaoft A programming language written with python How to use A simple Hello World: c="Hello World" c Output: "Hello World" Operators: a=12

1 Jan 24, 2022
Source Code for our paper: Understand me, if you refer to Aspect Knowledge: Knowledge-aware Gated Recurrent Memory Network

KaGRMN-DSG_ABSA This repository contains the PyTorch source Code for our paper: Understand me, if you refer to Aspect Knowledge: Knowledge-aware Gated

XingBowen 4 May 20, 2022
Pytorch implementation of OCNet series and SegFix.

openseg.pytorch News 2021/09/14 MMSegmentation has supported our ISANet and refer to ISANet for more details. 2021/08/13 We have released the implemen

openseg-group 1.1k Dec 23, 2022
Speech Recognition is an important feature in several applications used such as home automation, artificial intelligence

Speech Recognition is an important feature in several applications used such as home automation, artificial intelligence, etc. This article aims to provide an introduction on how to make use of the S

RISHABH MISHRA 1 Feb 13, 2022
Hand gesture recognition model that can be used as a remote control for a smart tv.

Gesture_recognition The training data consists of a few hundred videos categorised into one of the five classes. Each video (typically 2-3 seconds lon

Pratyush Negi 1 Aug 11, 2022
A clear, concise, simple yet powerful and efficient API for deep learning.

The Gluon API Specification The Gluon API specification is an effort to improve speed, flexibility, and accessibility of deep learning technology for

Gluon API 2.3k Dec 17, 2022
JAX + dataclasses

jax_dataclasses jax_dataclasses provides a wrapper around dataclasses.dataclass for use in JAX, which enables automatic support for: Pytree registrati

Brent Yi 35 Dec 21, 2022
LV-BERT: Exploiting Layer Variety for BERT (Findings of ACL 2021)

LV-BERT Introduction In this repo, we introduce LV-BERT by exploiting layer variety for BERT. For detailed description and experimental results, pleas

Weihao Yu 14 Aug 24, 2022
Parameter-ensemble-differential-evolution - Shows how to do parameter ensembling using differential evolution.

Ensembling parameters with differential evolution This repository shows how to ensemble parameters of two trained neural networks using differential e

Sayak Paul 9 May 04, 2022
Implementing yolov4 target detection and tracking based on nao robot

Implementing yolov4 target detection and tracking based on nao robot

6 Apr 19, 2022
A Real-ESRGAN equipped Colab notebook for CLIP Guided Diffusion

#360Diffusion automatically upscales your CLIP Guided Diffusion outputs using Real-ESRGAN. Latest Update: Alpha 1.61 [Main Branch] - 01/11/22 Layout a

78 Nov 02, 2022
PyTorch Implementation for "ForkGAN with SIngle Rainy NIght Images: Leveraging the RumiGAN to See into the Rainy Night"

ForkGAN with Single Rainy Night Images: Leveraging the RumiGAN to See into the Rainy Night By Seri Lee, Department of Engineering, Seoul National Univ

Seri Lee 52 Oct 12, 2022
Pytorch Implementation of Google's Parallel Tacotron 2: A Non-Autoregressive Neural TTS Model with Differentiable Duration Modeling

Parallel Tacotron2 Pytorch Implementation of Google's Parallel Tacotron 2: A Non-Autoregressive Neural TTS Model with Differentiable Duration Modeling

Keon Lee 170 Dec 27, 2022
RADIal is available now! Check the download section

Latest news: RADIal is available now! Check the download section. However, because we are currently working on the data anonymization, we provide for

valeo.ai 55 Jan 03, 2023