karolpiczak / esc-50 Goto Github PK

View Code? Open in Web Editor NEW

1.3K 31.0 284.0 878.79 MB

ESC-50: Dataset for Environmental Sound Classification

License: Other

Python 100.00%

dataset environmental-sounds audio

esc-50's Introduction

esc-50's People

Contributors

Stargazers

Watchers

Forkers

ankitshah009 xufuou nikolaypavlov shengkuizhao nature0310 felipebetancur yanbinchen panuganti abecadel ahoyosid radachynskaja ddoe777 suzhidong aascode maggie0830 marc-moreaux lnjume eranweissensternsc nemocpp muhammadmuhlas wantongtang usccolumbia fanwei918 lvaleriu thpatty zhaoforever runngezhang afcarl caiweicaiwei dreamerdoremi tangsg mikesneider calebebrim kpreetsid fitrialif robinjie peace098beat erayon aaqibrmir fl2016 twistedmove arunodhayan xinkez manik0007 xdcesc n130557 jeffw6 ramasubbaiya arjunmurthys singh-sidharth datajan02 idomatheveryday aquietlife baoqianyue xubo245 haotianchi josefranciscodiez contribu panandicoding ashishpatel26 appletree123123 qlongdevdn dengjincs zoltanszalontay lymgoforit fkqw yongyug licj1 xiongmaoxia readsharemvp templeblock xzm2004260 amoliu baucheng yanchang-liang kholondon hadryan wangpy1204 chaoxu1203 sergiperezc taku990 zui-jiang veyselkoparal sarathchandra-chen mercy18xmx gaojialu chalisthe mdamo savva kingtsong drericebert pragyasinghrathore gizachewbi ikhsanrahman khalilcharfi nonday manojkl madatonline bmartin1 akktan

esc-50's Issues

How to track particular sound class data

The data set consist of 2000 audio sounds, but i am interested in sounds classes like traffic sounds, helicopter making sounds, human sounds. how to track that particular classes.
tha data is not arranges as classes, it is mixed.

Can't download the E50 data

The download is always failing when it reaches ~510 MB. It says network error, EOF found something like that. Anyone else facing this issue? I have tried like 50 times to download it. With chrome browser as well as download managers.

ESC-50 results with audiovisual self-supervised learning

Hi Karolpiczak,
Thanks for organizing this repo for facilitating the development on sound event reg. In 2019, we have published one work @ CVPR2019, which is the first work that suppresses the human accuracy with audiovisual self-supervised learning approach, i.e., < Deep Multimodal Clustering for Unsupervised Audiovisual Learning>. Here is the link: http://openaccess.thecvf.com/content_CVPR_2019/html/Hu_Deep_Multimodal_Clustering_for_Unsupervised_Audiovisual_Learning_CVPR_2019_paper.html

We do hope that you could append our previous work in your next update.

Best,
Di

Sample 1-13572-A-46.wav

I think the audio sample 1-13572-A-46.wav might have different format with other clips.

Seperately categorize cough and sneeze

Is there any ways to seperately categorize only cough and sneeze sounds to train? them I am using pytorch to train datasets. Do I have to categorize using pandas?

ESC on Android

Is there a better ESC project on the android platform?

Appreciate you!

Wondering how spectrogram is drawn in the Readme

I've tried to reprduce it with python using matplotlib.pyplot.specgram, but can't get exactly the same figure, which seems to be a colormap problem. Great thanks if someone knows and tells me.

ESC-US subset

Hello Karol,
I'm using another related dataset ESC-US and I just wanted to confirm that there isn't any overlap between the files in ESC-50 and the files in ESC-US? The paper seems to imply that they are separate but I wanted to ask for confirmation.

How to divide the dataset into train/val/test?

Hi, I'd like to know how to divide the dataset into train/val/test. It seems to be divided as train/val=8:1. Is this true?

Created script to read the dataset into TensorFlow

Those searching on how to quickly use the dataset with TensorFlow can find an implementation here: esc50-tensorflow

License of ESC-US

Hi, thanks for sharing this amazing work.

I'd like to use ESC-US for my demo (which could be commercial) but couldn't find the license. Could you please let me know which license is applied to ESC-US?

Thanks!