Open AccessReal-Time Person Segmentation – Based on Body Pix
Author(s) -
Anish Mankotia and Meenu Garg
Publication year - 2020
Publication title -
international journal of modern trends in science and technology
Language(s) - English
Resource type - Journals
ISSN - 2455-3778
DOI - 10.46501/ijmtst061201
Subject(s) - segmentation , computer science , workflow , convolutional neural network , computation , artificial intelligence , class (philosophy) , tensor (intrinsic definition) , computer vision , pattern recognition (psychology) , algorithm , mathematics , database , pure mathematics
In this paper, we propose a novel semantic segmentation-based on the body pix module of the Tensor flow.jswhich can keep up with the accuracy of the state-of-the art approaches while running in real time. Thesolution follows the convolutional neural networks, each step in the workflow being enhanced by additionalinformation from semantic segmentation. Therefore, we introduce several improvements to computation,aggregation, and optimization by adapting existing techniques to integrate additional surface informationgiven by each semantic class. Using the body pix model which is trained using CNN, the ResNET50, thisnetwork can work with more than 150 layers, removing the problem of vanishing gradients. Using thisnetwork our body pix module, creates a more accurate and defined segmentation, and also supportsmulti-person segmentation.