11:45 am, Lecture Hall 5
MSc Thesis Presentation
People detection - An SVD-SVM Approach
Chennai Mathematical Institute.
Detecting humans is crucial for diverse application areas, including but not limited to abnormal event detection, congestion analysis, person counting, person identification, gender classification. As such it becomes useful that this topic is studied deeply. In the talk, we tackle one of the problems - the problem on person counting in a very specific setting. We introduce machine learning paradigm. We then move on to the current available techniques, discuss some of the approaches tried by us and their respective results. We then move on to discuss the algorithm based largely on concepts of Singular Value decomposition. We then explore some concepts of Support vector machines and the result of the algorithm developed. We then discuss neural networks which might be useful in taking the research forward.
The talk will be accessible to all. The only pre-requisite for the talk is keenness to see how machine learning paradigms work in real world settings.