PROJECT TITLE :
Visual Workflow Recognition Using a Variational Bayesian Treatment of Multistream Fused Hidden Markov Models
In this paper, we offer a variational Bayesian (VB) treatment of multistream fused hidden Markov models (MFHMMs), and apply it within the context of active learning-based visual workflow recognition (WR). Contrary to coaching strategies yielding point estimates, like most likelihood or maximum a posteriori coaching, the VB approach provides an estimate of the posterior distribution over the MFHMM parameters. Therefore, our approach provides a sublime resolution toward the amelioration of the overfitting problems of point estimate-based ways. Additionally, it provides a measure of confidence within the accuracy of the learned model, therefore permitting for the straightforward and cost-effective utilization of active learning in the context of MFHMMs. Two different active learning algorithms are considered during this paper: query by committee, that selects unlabeled data that minimize the classification variance, and a maximum information gain methodology that aims to maximize the alteration in model variance by correct data labeling. We tend to demonstrate the efficacy of the proposed treatment of MFHMMs by examining two difficult WR eventualities, and show that the appliance of active learning, which is facilitated by our VB approach, permits for a vital reduction of the MFHMM training costs.
Did you like this research project?
To get this research project Guidelines, Training and Code... Click Here