DCASE 2023 Challenge Task 7

Keunwoo Choi; Jaekwon Im; Laurie M. Heller; Brian McFee; Keisuke Imoto; Yuki Okamoto; Mathieu Lagrange; Shinnosuke Takamichi

doi:10.48550/arXiv:2304.12521

Published June 28, 2023 | Version v1

Video/Audio Open

DCASE 2023 Challenge Task 7

1. Gaudio Lab, inc.
2. Gaudio Lab, inc. / KAIST
3. Carnegie Mellon University
4. New York University
5. Doshisha University
6. Ritsumeikan University
7. CNRS, Ecole Centrale Nantes, Nantes Université
8. The University of Tokyo

Description

These audio files are the dataset and generated sound samples from the submission systems for the DCASE 2023 Challenge Task 7 "Foley Sound Synthesis. The audio data consists of sound excerpts which are divided into 7 classes. Each sound excerpt is a mono 16-bit 22,050 Hz 4-second audio. The classes of audio are as follows.

DogBark
Footstep
GunShot
Keyboard
MovingMotorVehicle
Rain
Sneeze/Cough

Each submission consists of 700 sound excerpts in each category. For more details, please check the challenge website.

A portion of the sounds in the data set were kindly provided by permission of the BBC under the condition that the sounds are used for this DCASE challenge and are only used for research purposes. You can check the original source of each sound in 'DevMeta.csv' and 'EvalMeta.csv' located in 'DCASE_2023_Challenge_Task_7_Dataset'.

Directory structure