Yoga Pose Classification Using Transfer Learning

TL;DR

Yoga-82, a benchmark dataset for large-scale yoga pose recognition with 82 classes, has challenging positions that could make precise annotations impossible, so VGG-16, ResNet-50, ResNet-101, and DenseNet-121 are used and finetuned them in different ways to get better results.

Abstract

Yoga has recently become an essential aspect of human existence for maintaining a healthy body and mind. People find it tough to devote time to the gym for workouts as their lives get more hectic and they work from home. This kind of human pose estimation is one of the notable problems as it has to deal with locating body key points or joints. Yoga-82, a benchmark dataset for large-scale yoga pose recognition with 82 classes, has challenging positions that could make precise annotations impossible. We have used VGG-16, ResNet-50, ResNet-101, and DenseNet-121 and finetuned them in different ways to get better results. We also used Neural Architecture Search to add more layers on top of this pre-trained architecture. The experimental result shows the best performance of DenseNet-121 having the top-1 accuracy of 85% and top-5 accuracy of 96% outperforming the current state-of-the-art result.

Figures (11)

• Figure 1: Our Proposed Workflow Diagram.
• Figure 2: Yoga-82 Dataset Description 1_verma2020yoga.
• Figure 3: Effect of the contrast factor.
• Figure 4: Effect of the median filter.
• Figure 5: Effect of the image sharpening.