The College of New Jersey

Apply     Visit     Give     |     Alumni     Parents     Offices     TCNJ Today     Three Bar Menu

Colloquium Talk with Honglu Zhou: April 1: Intelligent Video Understanding through Relational and Compositional Reasoning

Honglu Zhou, Ph.D. student at Rutgers University in the Department of Computer Science, will give a virtual colloquium talk on Friday, April 1, from 12:30 – 1:30 PM.  Honglu will share her research projects in machine learning applications in computer vision and graphics.

See below for more information about Honglu Zhou and the links for the event.

Abstract: Our experience as humans is deeply shaped by our perception of what happens to the objects in the visual world. Rather than building a machine that attempts to attain visual intelligence from the static and low-level pixels of images, we might need to accomplish the non-trivial higher-level visual understanding from object-centric learning of videos. Among a few critical directions for visual perception and machine intelligence, relational reasoning that reasons the saliency of objects and their dynamic interactions, and compositional learning where we compose and decompose symbolic objects in order to form holistic representations can help us develop robust and generalizable systems that can not only visually perceive but also understand and even interact with the world. In this talk, I will introduce our work on relational reasoning and compositional learning of videos.

Speaker Bio: Honglu Zhou is a Ph.D. student at Rutgers University in the Department of Computer Science, under the supervision of Prof. Mubbasir Kapadia. Her research interests mainly lie in Computer Vision and Deep Learning. She is passionate about the next-generation machine intelligence, especially machine learning and machine reasoning that enable a deeper understanding of the semantics of real-world data, which can be in forms of video, graph, human skeleton and many more. Projects that she has been working on include human group activity recognition from videos, video chapter generation, spatiotemporal reasoning and object tracking, predicting crowd dynamics, enabling intelligent and automatic floorplan design, forecasting online information spread, etc. She is currently researching on how to augment deep neural networks with relational and compositional reasoning capabilities to enrich a higher level computational video understanding.

Zoom Meeting (ID: 957 7840 7919 / Password: 464063)