Brian Chen

I’m a senior researcher at Samsung Research America working on video summarization and text-to-image editing/style transfer. Before I joined SRA, I was a visiting researcher at Meta Reality Lab/FAIR Embodied AI. I received my Ph.D. at Dept. Of Computer Science, Columbia University, in DVMM lab advised by Prof. Shih-Fu Chang.

My research interests focus on Computer Vision, Multimodal Learning, and Self-supervised Learning. Particularly, I am interested in learning representations from videos.

In Meta, I worked on a project to develop egocentric agents on Aria that understand everyday tasks specified in natural language and generate goal plans for further assistance. We leverage the large language model (LLM) LLaMA2 and video training for zero-shot goal planning. Our model can generate future plans conditioned on the given video and input goals.

During my Ph.D. period, I worked on the DARPA AIDA project, which mainly focused on incorporating cross-domain knowledge (images, videos, text, and audio) for knowledge graph construction.

I am closely working with IBM Research and MIT CSAIL on the Sight and Sound Project (since 2020 - ), aiming at learning representations from video and audio.

Prior to joining Columbia Univ., I finished my Bachelor and Master degrees at the Dept. of Computer Science and Information Eng., National Taiwan University, in 2015 and 2017 respectively, advised by Prof. Shou-De Lin.