Bayesian Multi-Object Tracking for Cell Microscopy

Abstract

Cell tracking is an essential tool for studying how cells behave and divide under different conditions. This thesis proposes new approaches to track cells and their lineages using random finite set, which allows the tracking errors to be statistically quantified. Additionally, this thesis also explores criteria to rank performance of basic vision task algorithms (e.g., object detection, instance-level segmentation, and tracking), which have not been received proportionate attention from the scientific community.