A dataset for evaluating visual-odometry algorithms in dynamic environments
Speaker
Sam Schofield
Institute
University of Canterbury
Time & Place
Mon, 02 Nov 2020 15:00:00 NZDT in E16 - Engineering Core
Abstract
Visual odometry (VO) is the process of estimating the motion of a camera using video that it has captured. A common assumption made by VO algorithms is that the environment is static. Existing datasets uphold this assumption, making it challenging to evaluate and improve the robustness of VO in dynamic environments. We are developing a dataset that provides challenging dynamic scenes and are using it to evaluate the performance of state-of-the-art visual odometry algorithms.
Biography
Sam is a PhD student focusing on improving the robustness of visual odometry in challenging environments. He is also a research assistant for UCDroneLab working on autonomous UAVs.