Chroma-Actions Dataset: Acoustic Images

Published: 10 January 2023| Version 1 | DOI: 10.17632/r4r4m2vjvh.1
Contributors:
Muhammad Bilal Shaikh,
,
,

Description

Chromagram-based representation of audio extracted from videos. These representations were extracted from the UCF-101 Human Action Recognition dataset. Only videos with audio channels were considered.

Files

Steps to reproduce

How the data were acquired Audios of human actions were extracted from UCF101, which was originally collected from YouTube. A script was devised to extract audios of actions from fifty-one different action categories: Archery, Cricket Shot, Hair Cutting, Playing Flute, Rafting, Sky Diving and so on. Data were arranged in two folders train and test to help researchers in evaluating their models.

Institutions

Edith Cowan University, University of Western Australia

Categories

Computer Vision Representation

Funding

Higher Education Commision, Pakistan

PM/HRDI-UESTPs/UETs- 456 I/Phase-1/Batch-VI/2018

License