Acoustic Event Dataset

 

Description

This data set contains 28 class acoustic event sounds.

 

Event class	Total length [min]	# sample
acoustic_guitar	23.4	190
airplane	37.9	198
applause	41.6	278
bird	46.3	265
car	38.5	231
cat	21.3	164
child	19.5	115
church_bell	11.8	71
crowd	64.6	328
dog_barking	9.2	113
engine	47.8	263
fireworks	43	271
footstep	70.3	378
glass_breaking	4.3	86
hammer	42.5	240
helicopter	22.1	111
knock	10.4	108
laughter	24.7	201
mouse_click	14.6	96
ocean_surf	42	218
rustle	22.8	184
scream	5.3	59
speech	18.9	99
squeak	19.8	173
tone	14.1	155
violin	16.1	162
water_tap	30.2	208
whistle	6	78
Total	768.4	5223

 

 

Policy

If you end up using the dataset, we ask you to cite the following paper:

 

Naoya Takahashi, Michael Gygli, Beat Pfister and Luc Van Gool,

"Deep Convolutional Neural Networks and Data Augmentation for Acoustic Event Recognition",

 Proc. Interspeech 2016, San Fransisco

 

 

Download

The dataset can be downloaded here. (1.2GB)

 

The code to run the pre-trained net can be found on bitbucket.