Predicting accuracy in Multiple Object Tracking tasks from trajectory statistics

Abstract

Title: Predicting accuracy in Multiple Object Tracking tasks from trajectory statistics Author: Surya Prakash Chembrolu Department: Department of Software and Computer Science Education Supervisor: Mgr. Filip Děchtěrenko, Ph.D., Department of Software and Com- puter Science Education Abstract: Cognitive science is an interdisciplinary area covering neuroscience, psychology, linguistics, philosophy, and computer science. Computer science and cognitive science mutually benefit from each other because computer science is very helpful to design and perform experiments in order to understand how the brain works likewise research output from cognitive science can lead to new con- cepts and models in artificial intelligence. Within cognitive science, Multiple Object Tracking (MOT) paradigm is used to study visual attention. In MOT experiments, participants are required to keep track of some moving objects in parallel. In this study, a data-driven approach is taken in order to explain the tracking performance of the subjects taking part in MOT experiments. The stimuli in MOT known as trajectories or tracks presented in previous studies were taken and the difficulty of those trajectories is quantified based on trajec- tory statistics. Then a model is created to explain tracking performance and this model is tested...