https://github.com/daniel-furman/esc-50

ESC-50: Dataset for Environmental Sound Classification
Science Score: 23.0%

This score indicates how likely this project is to be science-related based on various indicators:
○
CITATION.cff file
○
codemeta.json file
○
.zenodo.json file
✓
DOI references
Found 4 DOI reference(s) in README
✓
Academic publication links
Links to: arxiv.org, researchgate.net, sciencedirect.com, acm.org
○
Academic email domains
○
Institutional organization owner
○
JOSS paper metadata
○
Scientific vocabulary similarity
Low similarity (8.0%) to scientific vocabulary
Last synced: 6 months ago · JSON representation
Repository

ESC-50: Dataset for Environmental Sound Classification
Basic Info

Host: GitHub
Owner: daniel-furman
License: other
Default Branch: master
Homepage:
Size: 858 MB
Statistics

Stars: 0
Watchers: 0
Forks: 0
Open Issues: 0
Releases: 0
Fork of karolpiczak/ESC-50
Created over 4 years ago · Last pushed over 4 years ago
https://github.com/daniel-furman/ESC-50/blob/master/

## ESC-50: Dataset for Environmental Sound Classification

> ###### [Overview](#esc-50-dataset-for-environmental-sound-classification) | [Download](#download) | [Results](#results) | [Repository content](#repository-content) | [License](#license) | [Citing](#citing) | [Caveats](#caveats) | [Changelog](#changelog)
>
>  
 
 



The **ESC-50 dataset** is a labeled collection of 2000 environmental audio recordings suitable for benchmarking methods of environmental sound classification.

The dataset consists of 5-second-long recordings organized into 50 semantical classes (with 40 examples per class) loosely arranged into 5 major categories:

| _Animals | _{Natural soundscapes & water sounds} | _{Human, non-speech sounds} | _{Interior/domestic sounds} | _{Exterior/urban noises} |
| :--- | :--- | :--- | :--- | :--- |
| _Dog | _Rain | _{Crying baby} | _{Door knock} | _Helicopter |
| _Rooster | _{Sea waves} | _Sneezing | _{Mouse click} | _Chainsaw |
| _Pig | _{Crackling fire} | _Clapping | _{Keyboard typing} | _Siren |
| _Cow | _Crickets | _Breathing | _{Door, wood creaks} | _{Car horn} |
| _Frog | _{Chirping birds} | _Coughing | _{Can opening} | _Engine |
| _Cat | _{Water drops} | _Footsteps | _{Washing machine} | _Train |
| _Hen | _Wind | _Laughing | _{Vacuum cleaner} | _{Church bells} |
| _{Insects (flying)} | _{Pouring water} | _{Brushing teeth} | _{Clock alarm} | _Airplane |
| _Sheep | _{Toilet flush} | _Snoring | _{Clock tick} | _Fireworks |
| _Crow | _Thunderstorm | _{Drinking, sipping} | _{Glass breaking} | _{Hand saw} |

Clips in this dataset have been manually extracted from public field recordings gathered by the **[Freesound.org project](http://freesound.org/)**. The dataset has been prearranged into 5 folds for comparable cross-validation, making sure that fragments from the same original source file are contained in a single fold.

A more thorough description of the dataset is available in the original [paper](http://karol.piczak.com/papers/Piczak2015-ESC-Dataset.pdf) with some supplementary materials on GitHub: **[ESC: Dataset for Environmental Sound Classification - paper replication data](https://github.com/karoldvl/paper-2015-esc-dataset)**.


## Download

The dataset can be downloaded as a single .zip file (~600 MB):

**[Download ESC-50 dataset](https://github.com/karoldvl/ESC-50/archive/master.zip)**


## Results

Numerous machine learning & signal processing approaches have been evaluated on the ESC-50 dataset. Most of them are listed here. If you know of some other reference, you can message me or open a Pull Request directly.

> ###### Terms used in the table:
> 
> _{CNN - Convolutional Neural Network
 CRNN - Convolutional Recurrent Neural Network
 GMM - Gaussian Mixture Model
 GTCC - Gammatone Cepstral Coefficients
 GTSC - Gammatone Spectral Coefficients
 k-NN - k-Neareast Neighbors
 MFCC - Mel-Frequency Cepstral Coefficients
 MLP - Multi-Layer Perceptron
 RBM - Restricted Boltzmann Machine
 RNN - Recurrent Neural Network
 SVM - Support Vector Machine
 TEO - Teager Energy Operator
 ZCR - Zero-Crossing Rate}

| _Title | _Notes | _Accuracy | _Paper | _Code |
| :--- | :--- | :--- | :--- | :--- |
| _{**AST: Audio Spectrogram Transformer**} | _{Pure Attention Model Pretrained on AudioSet} | _95.70% | _{[gong2021](https://arxiv.org/pdf/2104.01778.pdf)} | :scroll: |
| _{**A Sequential Self Teaching Approach for Improving Generalization in Sound Event Recognition**} | _{Multi-stage sequential learning with knowledge transfer from Audioset} | _94.10% | _{[kumar2020](https://arxiv.org/pdf/2007.00144.pdf)} |  |
| _{**Urban Sound Tagging using Multi-Channel Audio Feature with Convolutional Neural Networks**} | _{Pretrained model with multi-channel features} | _89.50% | _{[kim2020](http://dcase.community/documents/challenge2020/technical_reports/DCASE2020_JHKim_21_t5.pdf)} | :scroll: |
| _{**An Ensemble of Convolutional Neural Networks for Audio Classification**} | _{CNN ensemble with data augmentation} | _88.65% | _{[nanni2020](https://arxiv.org/pdf/2007.07966.pdf)} | :scroll: |
| _{**Environmental Sound Classification on the Edge: A Pipeline for Deep Acoustic Networks on Extremely Resource-Constrained Devices**} | _{CNN model (ACDNet) with potential compression} | _87.1% | _{[mohaimenuzzaman2021](https://arxiv.org/pdf/2103.03483.pdf)} | :scroll: |
| _{**Unsupervised Filterbank Learning Using Convolutional Restricted Boltzmann Machine for Environmental Sound Classification**} | _{CNN with filterbanks learned using convolutional RBM + fusion with GTSC and mel energies} | _86.50% | _{[sailor2017](https://pdfs.semanticscholar.org/f6fd/1be38a2d764d900b11b382a379efe88b3ed6.pdf)} |  |
| _{**AclNet: efficient end-to-end audio classification CNN**} | _{CNN with mixup and data augmentation} | _85.65% | _{[huang2018](https://arxiv.org/pdf/1811.06669.pdf)} |  |
| _{**On Open-Set Classification with L3-Net Embeddings for Machine Listening Applications**} | _{x-vector network with openll3 embeddings} | _85.00% | _{[wilkinghoff2020](https://www.eurasip.org/Proceedings/Eusipco/Eusipco2020/pdfs/0000800.pdf)} |  |
| _{**Learning from Between-class Examples for Deep Sound Recognition**} | _{EnvNet-v2 ([tokozume2017a](http://www.mi.t.u-tokyo.ac.jp/assets/publication/LEARNING_ENVIRONMENTAL_SOUNDS_WITH_END-TO-END_CONVOLUTIONAL_NEURAL_NETWORK-poster.pdf)) + data augmentation + Between-Class learning} | _84.90% | _{[tokozume2017b](https://openreview.net/forum?id=B1Gi6LeRZ)} |  |
| _{**Novel Phase Encoded Mel Filterbank Energies for Environmental Sound Classification**} | _{CNN working with phase encoded mel filterbank energies (PEFBEs), fusion with Mel energies} | _84.15% | _{[tak2017](https://www.researchgate.net/profile/Dharmesh_Agrawal/publication/320733074_Novel_Phase_Encoded_Mel_Filterbank_Energies_for_Environmental_Sound_Classification/links/5a084c780f7e9b68229c8947/Novel-Phase-Encoded-Mel-Filterbank-Energies-for-Environmental-Sound-Classification.pdf)} |  |
| _{**Knowledge Transfer from Weakly Labeled Audio using Convolutional Neural Network for Sound Events and Scenes**} | _{CNN pretrained on AudioSet} | _83.50% | _{[kumar2017](https://arxiv.org/pdf/1711.01369.pdf)} | :scroll: |
| _{**Unsupervised Filterbank Learning Using Convolutional Restricted Boltzmann Machine for Environmental Sound Classification**} | _{CNN with filterbanks learned using convolutional RBM + fusion with GTSC} | _83.00% | _{[sailor2017](https://pdfs.semanticscholar.org/f6fd/1be38a2d764d900b11b382a379efe88b3ed6.pdf)} |  |
| _{**Deep Multimodal Clustering for Unsupervised Audiovisual Learning**} | _{CNN + unsupervised audio-visual learning} | _82.60% | _{[hu2019](http://openaccess.thecvf.com/content_CVPR_2019/papers/Hu_Deep_Multimodal_Clustering_for_Unsupervised_Audiovisual_Learning_CVPR_2019_paper.pdf)} |  |
| _{**Novel TEO-based Gammatone Features for Environmental Sound Classification**} | _{Fusion of GTSC & TEO-GTSC with CNN} | _81.95% | _{[agrawal2017](http://www.eurasip.org/Proceedings/Eusipco/Eusipco2017/papers/1570347591.pdf)} |  |
| _{**Learning from Between-class Examples for Deep Sound Recognition**} | _{EnvNet-v2 ([tokozume2017a](http://www.mi.t.u-tokyo.ac.jp/assets/publication/LEARNING_ENVIRONMENTAL_SOUNDS_WITH_END-TO-END_CONVOLUTIONAL_NEURAL_NETWORK-poster.pdf)) + Between-Class learning} | _81.80% | _{[tokozume2017b](https://openreview.net/forum?id=B1Gi6LeRZ)} |  |
| :headphones: _{***Human accuracy***} | _{Crowdsourcing experiment in classifying ESC-50 by human listeners} | _81.30% | _{[piczak2015a](http://karol.piczak.com/papers/Piczak2015-ESC-Dataset.pdf)} | :scroll: |
| _{**Objects that Sound**} | _{*Look, Listen and Learn* (L3) network ([arandjelovic2017a](https://arxiv.org/pdf/1705.08168.pdf)) with stride 2, larger batches and learning rate schedule} | _79.80% | _{[arandjelovic2017b](https://arxiv.org/pdf/1712.06651.pdf)} |  |
| _{**Look, Listen and Learn**} | _{8-layer convolutional subnetwork pretrained on an audio-visual correspondence task} | _79.30% | _{[arandjelovic2017a](https://arxiv.org/pdf/1705.08168.pdf)} |  |
| _{**Learning Environmental Sounds with Multi-scale Convolutional Neural Network**} | _{Multi-scale convolutions with feature fusion (waveform + spectrogram)} | _79.10% | _{[zhu2018](https://arxiv.org/pdf/1803.10219.pdf)} |  |
| _{**Novel TEO-based Gammatone Features for Environmental Sound Classification**} | _{GTSC with CNN} | _79.10% | _{[agrawal2017](http://www.eurasip.org/Proceedings/Eusipco/Eusipco2017/papers/1570347591.pdf)} |  |
| _{**Learning from Between-class Examples for Deep Sound Recognition**} | _{EnvNet-v2 ([tokozume2017a](http://www.mi.t.u-tokyo.ac.jp/assets/publication/LEARNING_ENVIRONMENTAL_SOUNDS_WITH_END-TO-END_CONVOLUTIONAL_NEURAL_NETWORK-poster.pdf)) + data augmentation} | _78.80% | _{[tokozume2017b](https://openreview.net/forum?id=B1Gi6LeRZ)} |  |
| _{**Unsupervised Filterbank Learning Using Convolutional Restricted Boltzmann Machine for Environmental Sound Classification**} | _{CNN with filterbanks learned using convolutional RBM} | _78.45% | _{[sailor2017](https://pdfs.semanticscholar.org/f6fd/1be38a2d764d900b11b382a379efe88b3ed6.pdf)} |  |
| _{**Learning from Between-class Examples for Deep Sound Recognition**} | _{Baseline CNN ([piczak2015b](http://karol.piczak.com/papers/Piczak2015-ESC-ConvNet.pdf)) + Batch Normalization + Between-Class learning} | _76.90% | _{[tokozume2017b](https://openreview.net/forum?id=B1Gi6LeRZ)} |  |
| _{**Novel TEO-based Gammatone Features for Environmental Sound Classification**} | _{TEO-GTSC with CNN} | _74.85% | _{[agrawal2017](http://www.eurasip.org/Proceedings/Eusipco/Eusipco2017/papers/1570347591.pdf)} |  |
| _{**Learning from Between-class Examples for Deep Sound Recognition**} | _{EnvNet-v2 ([tokozume2017a](http://www.mi.t.u-tokyo.ac.jp/assets/publication/LEARNING_ENVIRONMENTAL_SOUNDS_WITH_END-TO-END_CONVOLUTIONAL_NEURAL_NETWORK-poster.pdf))} | _74.40% | _{[tokozume2017b](https://openreview.net/forum?id=B1Gi6LeRZ)} |  |
| _{**Soundnet: Learning sound representations from unlabeled video**} | _{8-layer CNN (raw audio) with transfer learning from unlabeled videos} | _74.20% | _{[aytar2016](http://papers.nips.cc/paper/6146-soundnet-learning-sound-representations-from-unlabeled-video.pdf)} | :scroll: |
| _{**Learning from Between-class Examples for Deep Sound Recognition**} | _{18-layer CNN on raw waveforms ([dai2016](https://arxiv.org/pdf/1610.00087.pdf)) + Between-Class learning} | _73.30% | _{[tokozume2017b](https://openreview.net/forum?id=B1Gi6LeRZ)} |  |
| _{**Novel Phase Encoded Mel Filterbank Energies for Environmental Sound Classification**} | _{CNN working with phase encoded mel filterbank energies (PEFBEs)} | _73.25% | _{[tak2017](https://www.researchgate.net/profile/Dharmesh_Agrawal/publication/320733074_Novel_Phase_Encoded_Mel_Filterbank_Energies_for_Environmental_Sound_Classification/links/5a084c780f7e9b68229c8947/Novel-Phase-Encoded-Mel-Filterbank-Energies-for-Environmental-Sound-Classification.pdf)} |  |
| _{**Classifying environmental sounds using image recognition networks**} | _{16 kHz sampling rate, GoogLeNet on spectrograms (40 ms frame length)} | _73.20% | _{[boddapati2017](https://www.sciencedirect.com/science/article/pii/S1877050917316599)} | :scroll: |
| _{**Learning from Between-class Examples for Deep Sound Recognition**} | _{Baseline CNN ([piczak2015b](http://karol.piczak.com/papers/Piczak2015-ESC-ConvNet.pdf)) + Batch Normalization} | _72.40% | _{[tokozume2017b](https://openreview.net/forum?id=B1Gi6LeRZ)} |  |
| _{**Novel TEO-based Gammatone Features for Environmental Sound Classification**} | _{Fusion of MFCC & TEO-GTCC with GMM} | _72.25% | _{[agrawal2017](http://www.eurasip.org/Proceedings/Eusipco/Eusipco2017/papers/1570347591.pdf)} |  |
| _{**Learning environmental sounds with end-to-end convolutional neural network (EnvNet)**} | _{Combination of spectrogram and raw waveform CNN} | _71.00% | _{[tokozume2017a](http://www.mi.t.u-tokyo.ac.jp/assets/publication/LEARNING_ENVIRONMENTAL_SOUNDS_WITH_END-TO-END_CONVOLUTIONAL_NEURAL_NETWORK-poster.pdf)} |  |
| _{**Novel TEO-based Gammatone Features for Environmental Sound Classification**} | _{TEO-GTCC with GMM} | _68.85% | _{[agrawal2017](http://www.eurasip.org/Proceedings/Eusipco/Eusipco2017/papers/1570347591.pdf)} |  |
| _{**Classifying environmental sounds using image recognition networks**} | _{16 kHz sampling rate, AlexNet on spectrograms (30 ms frame length)} | _68.70% | _{[boddapati2017](https://www.sciencedirect.com/science/article/pii/S1877050917316599)} | :scroll: |
| _{**Very Deep Convolutional Neural Networks for Raw Waveforms**} | _{18-layer CNN on raw waveforms} | _68.50% | _{[dai2016](https://arxiv.org/pdf/1610.00087.pdf), [tokozume2017b](https://openreview.net/forum?id=B1Gi6LeRZ)} | :scroll: |
| _{**Classifying environmental sounds using image recognition networks**} | _{32 kHz sampling rate, GoogLeNet on spectrograms (30 ms frame length)} | _67.80% | _{[boddapati2017](https://www.sciencedirect.com/science/article/pii/S1877050917316599)} | :scroll: |
| _{**WSNet: Learning Compact and Efficient Networks with Weight Sampling**} | _{SoundNet 8-layer CNN architecture with 100x model compression} | _66.25% | _{[jin2017](https://openreview.net/forum?id=H1I3M7Z0b)} |  |
| _{**Soundnet: Learning sound representations from unlabeled video**} | _{5-layer CNN (raw audio) with transfer learning from unlabeled videos} | _66.10% | _{[aytar2016](http://papers.nips.cc/paper/6146-soundnet-learning-sound-representations-from-unlabeled-video.pdf)} | :scroll: |
| _{**WSNet: Learning Compact and Efficient Networks with Weight Sampling**} | _{SoundNet 8-layer CNN architecture with 180x model compression} | _65.80% | _{[jin2017](https://openreview.net/forum?id=H1I3M7Z0b)} |  |
| _{**Soundnet: Learning sound representations from unlabeled video**} | _{5-layer CNN trained on raw audio of ESC-50 only} | _65.00% | _{[aytar2016](http://papers.nips.cc/paper/6146-soundnet-learning-sound-representations-from-unlabeled-video.pdf)} | :scroll: |
| _{:bar_chart: **Environmental Sound Classification with Convolutional Neural Networks** - ***CNN baseline***} | _{CNN with 2 convolutional and 2 fully-connected layers, mel-spectrograms as input, vertical filters in the first layer} | _64.50% | _{[piczak2015b](http://karol.piczak.com/papers/Piczak2015-ESC-ConvNet.pdf)} | :scroll: |
| _{**auDeep: Unsupervised Learning of Representations from Audio with Deep Recurrent Neural Networks**} | _{MLP classifier on features extracted with an RNN autoencoder} | _64.30% | _{[freitag2017](https://arxiv.org/pdf/1712.04382.pdf)} | :scroll: |
| _{**Classifying environmental sounds using image recognition networks**} | _{32 kHz sampling rate, AlexNet on spectrograms (30 ms frame length)} | _63.20% | _{[boddapati2017](https://www.sciencedirect.com/science/article/pii/S1877050917316599)} | :scroll: |
| _{**Classifying environmental sounds using image recognition networks**} | _CRNN | _60.30% | _{[boddapati2017](https://www.sciencedirect.com/science/article/pii/S1877050917316599)} | :scroll: |
| _{**Comparison of Time-Frequency Representations for Environmental Sound Classification using Convolutional Neural Networks**} | _{3-layer CNN with vertical filters on wideband mel-STFT (*median accuracy*)} | _*56.37%* | _{[huzaifah2017](https://arxiv.org/pdf/1706.07156.pdf)} |  |
| _{**Comparison of Time-Frequency Representations for Environmental Sound Classification using Convolutional Neural Networks**} | _{3-layer CNN with square filters on wideband mel-STFT (*median accuracy*)} | _*54.00%* | _{[huzaifah2017](https://arxiv.org/pdf/1706.07156.pdf)} |  |
| _{**Soundnet: Learning sound representations from unlabeled video**} | _{8-layer CNN trained on raw audio of ESC-50 only} | _51.10% | _{[aytar2016](http://papers.nips.cc/paper/6146-soundnet-learning-sound-representations-from-unlabeled-video.pdf)} | :scroll: |
| _{**Comparison of Time-Frequency Representations for Environmental Sound Classification using Convolutional Neural Networks**} | _{5-layer CNN with square filters on wideband mel-STFT (*median accuracy*)} | _*50.87%* | _{[huzaifah2017](https://arxiv.org/pdf/1706.07156.pdf)} |  |
| _{**Comparison of Time-Frequency Representations for Environmental Sound Classification using Convolutional Neural Networks**} | _{5-layer CNN with vertical filters on wideband mel-STFT (*median accuracy*)} | _*46.25%* | _{[huzaifah2017](https://arxiv.org/pdf/1706.07156.pdf)} |  |
| :bar_chart: _{***Baseline - random forest***} | _{Baseline ML approach (MFCC & ZCR + random forest)} | _44.30% | _{[piczak2015a](http://karol.piczak.com/papers/Piczak2015-ESC-Dataset.pdf)} | :scroll: |
| _{**Soundnet: Learning sound representations from unlabeled video**} | _{Convolutional autoencoder trained on unlabeled videos} | _39.90% | _{[aytar2016](http://papers.nips.cc/paper/6146-soundnet-learning-sound-representations-from-unlabeled-video.pdf)} | :scroll: |
| :bar_chart: _{***Baseline - SVM***} | _{Baseline ML approach (MFCC & ZCR + SVM)} | _39.60% | _{[piczak2015a](http://karol.piczak.com/papers/Piczak2015-ESC-Dataset.pdf)} | :scroll: |
| :bar_chart: _{***Baseline - k-NN***} | _{Baseline ML approach (MFCC & ZCR + k-NN)} | _32.20% | _{[piczak2015a](http://karol.piczak.com/papers/Piczak2015-ESC-Dataset.pdf)} | :scroll: |
| _{**A mixture model-based real-time audio sources classification method**} | _{Dictionary of sound models used for classification (*accuracy is computed on segments instead of files*)} | _*94.00%* | _{[baelde2017](https://hal.archives-ouvertes.fr/hal-01420677v2/document)} |  |
| _{**NELS - Never-Ending Learner of Sounds**} | _{Large-scale audio crawling with classifiers trained on AED datasets (including ESC-50)} | _N/A | _{[elizalde2017](http://media.aau.dk/smc/wp-content/uploads/2017/12/ML4AudioNIPS17_paper_3.pdf)} | :scroll: |
| _{**Utilizing Domain Knowledge in End-to-End Audio Processing**} | _{End-to-end CNN with learned mel-spectrogram transformation} | _N/A | _{[tax2017](https://arxiv.org/pdf/1712.00254.pdf)} | :scroll: |
| _{**Deep Neural Network based learning and transferring mid-level audio features for acoustic scene classification**} | _{Transfer learning from various datasets, including ESC-50} | _N/A | _{[mun2017](https://pdfs.semanticscholar.org/5f1e/513aec29b8c471723172f1d5bc2174daa71a.pdf)} |  |
| _{**Features and Kernels for Audio Event Recognition**} | _{MFCC, GMM, SVM} | _N/A | _{[kumar2016b](https://arxiv.org/pdf/1607.05765.pdf)} |  |
| _{**A real-time environmental sound recognition system for the Android OS**} | _{Real-time sound recognition for Android evaluated on ESC-10} | _N/A | _{[pillos2016](https://www.cs.tut.fi/sgn/arg/dcase2016/documents/workshop/Pillos-DCASE2016workshop.pdf)} |  |
| _{**Comparing Time and Frequency Domain for Audio Event Recognition Using Deep Learning**} | _{Discriminatory effectiveness of different signal representations compared on ESC-10 and Freiburg-106} | _N/A | _{[hertel2016](https://arxiv.org/pdf/1603.05824.pdf)} |  |
| _{**Audio Event and Scene Recognition: A Unified Approach using Strongly and Weakly Labeled Data**} | _{Combination of weakly labeled data (YouTube) with strong labeling (ESC-10) for Acoustic Event Detection} | _N/A | _{[kumar2016a](https://arxiv.org/pdf/1611.04871.pdf)} |  |


## Repository content

- [`audio/*.wav`](audio/)

  2000 audio recordings in WAV format (5 seconds, 44.1 kHz, mono) with the following naming convention:
  
  `{FOLD}-{CLIP_ID}-{TAKE}-{TARGET}.wav`
  
  - `{FOLD}` - index of the cross-validation fold,
  - `{CLIP_ID}` - ID of the original Freesound clip,
  - `{TAKE}` - letter disambiguating between different fragments from the same Freesound clip,
  - `{TARGET}` - class in numeric format [0, 49].

- [`meta/esc50.csv`](meta/esc50.csv)

  CSV file with the following structure:
  
  | _filename | _fold | _target | _category | _esc10 | _{src_file} | _take |
  | :--- | :--- | :--- | :--- | :--- | :--- | :--- |
  
  The `esc10` column indicates if a given file belongs to the *ESC-10* subset (10 selected classes, CC BY license).
  
- [`meta/esc50-human.xlsx`](meta/esc50-human.xlsx)

  Additional data pertaining to the crowdsourcing experiment (human classification accuracy).


## License

The dataset is available under the terms of the [Creative Commons Attribution Non-Commercial license](http://creativecommons.org/licenses/by-nc/3.0/).

A smaller subset (clips tagged as *ESC-10*) is distributed under CC BY (Attribution).

Attributions for each clip are available in the [ LICENSE file](LICENSE).


## Citing



If you find this dataset useful in an academic setting please cite:

> K. J. Piczak. **ESC: Dataset for Environmental Sound Classification**. *Proceedings of the 23rd Annual ACM Conference on Multimedia*, Brisbane, Australia, 2015.
> 
> [DOI: http://dx.doi.org/10.1145/2733373.2806390]

    @inproceedings{piczak2015dataset,
      title = {{ESC}: {Dataset} for {Environmental Sound Classification}},
      author = {Piczak, Karol J.},
      booktitle = {Proceedings of the 23rd {Annual ACM Conference} on {Multimedia}},
      date = {2015-10-13},
      url = {http://dl.acm.org/citation.cfm?doid=2733373.2806390},
      doi = {10.1145/2733373.2806390},
      location = {{Brisbane, Australia}},
      isbn = {978-1-4503-3459-4},
      publisher = {{ACM Press}},
      pages = {1015--1018}
    }

## Caveats

Please be aware of potential information leakage while training models on *ESC-50*, as some of the original Freesound recordings were already preprocessed in a manner that might be class dependent (mostly bandlimiting). Unfortunately, this issue went unnoticed when creating the original version of the dataset. Due to the number of methods already evaluated on *ESC-50*, no changes rectifying this issue will be made in order to preserve comparability.


## Changelog

###### v2.0.0 (2017-12-13)

> _{Change to WAV version as default.}

###### v2.0.0-pre (2016-10-10) (wav-files branch)

> _{Replace OGG recordings with cropped WAV files for easier loading and frame-level precision (some of the OGG recordings had a slightly different length when loaded).
 Move recordings to a one directory structure with a meta CSV file.}

###### v1.0.0 (2015-04-15)

> _{Initial version of the dataset (OGG format).}