Yoga for all: A Comprehensive Collection of Yoga Images and Videos dataset

Published: 21 April 2023| Version 1 | DOI: 10.17632/jc4mmnvcdk.1


The dataset comprises both images and videos depicting right and wrong postures for a variety of Yoga asanas. The focus of the dataset is on 10 specific Yoga postures, namely Anantasana, Ardhakati Chakrasana, Bhujangasana, Kati Chakrasana, Marjariasana, Parvatasana, Sarvangasana, Tadasana, Vajrasana, and Viparita Karani. The Image dataset comprises a total of 11,344 compressed images (256 × 256 dimension) and is organized into 10 subfolders, each corresponding to a specific Yoga asana. Within each subfolder, there are two additional folders labeled "Right Steps" and "Wrong Steps". The "Right Steps" folder contains several subfolders, each representing a specific step in the right sequence of the Yoga asana and displaying the corresponding images. On the other hand, the "Wrong Steps" folder includes multiple subfolders, each showing images of an wrong steps in the sequence of the Yoga asana. The Yoga asana video dataset consists of 8 videos for each posture, comprising 4 videos demonstrating the right posture from 4 different angles and 4 videos exhibiting the wrong posture from 4 different angles. This dataset includes a total of 80 compressed videos for 10 Yoga asanas, with 40 videos demonstrating the right postures captured from 4 different angles, and 40 videos illustrating the wrong postures from 4 different angles. The dataset has advantages for various groups, such as app developers, machine learning researchers, Yoga instructors, and Yoga practitioners. Machine learning researchers can utilize this dataset to train computer vision algorithms in recognizing and categorizing various yoga postures automatically. App developers can use the dataset to generate yoga apps that present users with visual guidance on executing each posture and keeping track of their progress.



Artificial Intelligence, Machine Learning, Yoga, Health, Posture Studies