Incorporating Short-Term Spatial-Temporal Information

IP1: Incorporating Short-Term Spatial-Temporal Information for Robotic Sensing

Lead: Chris McCool

PhD student:

In this project, we will incorporate short-term spatial-temporal information to improve robotic sensing. This will integrate information regarding the spatial shift of the robot/camera (how far it has moved) along with the estimated scene structure (depth/3D information) in conjunction with the temporal information (time between captured images). In particular, we are interested with “how to best incorporate spatial-temporal information explicitly within deep learnt models”. To address this objective research topics of interest include, how to:

    • rapidly generate data to train spatial-temporal models (WP1);
    • transform the spatial-temporal for use within deep learnt models and/or redesign networks to best incorporate spatial-temporal; and
    • estimate and make use of the uncertainty of the spatial-temporal information.