Conventional deep CNN's are used nowadays to solve complex problems that has a huge dataset. Transfer learning is a new paradigm where knowledge gained from other models can be incorporated into other deep learning framework to produce better results. In this study, we use the pre-trained deep convolutional neural network models VGG16, ResNet50 and Xception to train our data on KB dataset that contains poses from karate - martial arts and bharathanatyam - classical dance. The performance accuracy acquired by each model are: VGG16 - 82.12, ResNet50 - 84.51 and Xception 87.34.
Volume 12 | 04-Special Issue
Pages: 973-977
DOI: 10.5373/JARDCS/V12SP4/20201569