ESC-50 Dataset
Load the ESC-50 dataset in Python with one line of code in seconds and plug it in TensorFlow and PyTorch with Activeloop Hub.
ESC-50 dataset
What is ESC-50 Dataset?
The ESC-50 (Environmental Sound Classification 50) Dataset is a collection of 2000 audio recordings of environment suitable for benchmarking methods of environmental sound classification. The dataset consists of 50 semantical classes with each having 5-second-long recordings and loosely organised into 5 major categories.
Download ESC-50 Dataset in Python
Instead of downloading the ESC-50 dataset in Python, you can effortlessly load it in Python via our open-source package Hub with just one line of code.
Load ESC-50 Dataset Training Subset in Python
1
import hub
2
ds = hub.load("hub://activeloop/esc50")
Copied!
ESC-50 Dataset Structure
ESC-50 Data Fields
- audio: tensor containing audio data
- labels: tensor containing labels for the audio
- esc10: tensor containing boolean values of whether the data contains in esc10 dataset
- take: tensor containing the revision number of the audio
- fold: tensor containing fold number of the audio
- target: tensor containing id of the audio
- src_file: tensor containing file name of the audio
How to use ESC-50 Dataset with PyTorch and TensorFlow in Python
Train a model on ESC-50 dataset with PyTorch in Python
1
dataloader = ds.pytorch(num_workers=0, batch_size=4, shuffle=False)
Copied!
Train a model on ESC-50 dataset with TensorFlow in Python
1
dataloader = ds.tensorflow()
Copied!
Additional Information about ESC-50 Dataset
ESC-50 Dataset Description
- Paper: Piczak, K. J. (2015, October). ESC: Dataset for environmental sound classification. In Proceedings of the 23rd ACM international conference on Multimedia (pp. 1015-1018).
ESC-50 Dataset Curators
K. J. Piczak
ESC-50 Dataset Licensing Information
ESC-50 Dataset Citation Information
1
@inproceedings{piczak2015dataset,
2
title = {{ESC}: {Dataset} for {Environmental Sound Classification}},
3
author = {Piczak, Karol J.},
4
booktitle = {Proceedings of the 23rd {Annual ACM Conference} on {Multimedia}},
5
date = {2015-10-13},
6
url = {http://dl.acm.org/citation.cfm?doid=2733373.2806390},
7
doi = {10.1145/2733373.2806390},
8
location = {{Brisbane, Australia}},
9
isbn = {978-1-4503-3459-4},
10
publisher = {{ACM Press}},
11
pages = {1015--1018}
12
}
Copied!
Last modified 27d ago