Zhiqi Li (PhD, FMC) with this poster entitled: Deep learning for scene flow estimation: methods and applications.
Click the poster below to enlarge.
Scene flow estimation aims at obtaining structure information and 3D motion of dynamic scenes. It has long been an interest of research in computer vision and 3D computer graphics. It is a fundamental task for various applications like autonomous driving. Compared to previous methods utilizing image representations, many recent researches build upon the power of deep analysis on point clouds and focus on point clouds representation to conduct 3D flow estimation. In this survey, we comprehensively review the pioneering literature in scene flow estimation based on point clouds, delve in detail into their learning paradigms and present insightful comparison between the state-of-the-art methods using deep learning for scene flow estimation. Furthermore, we introduce various higher-level scene understanding tasks (object tracking, motion segmentation, etc.) which could benefit from the latest progress on scene flow estimation. The paper concludes with an overview of foreseeable research trends for scene flow estimation.
You can view the full poster exhibition and pre-recorded presentations on the conference webpage.