Hao Chen
陈 豪
haochen.umd@gmail.com


I am currently a Research Scientist at ByteDance. Before this, I worked as a Postdoctoral Researcher at Meta. I earned my Ph.D. in Computer Science from the University of Maryland, College Park, under the guidance of Prof. Abhinav Shrivastava. Prior to this, I completed my Master's degree in Pattern Recognition & Intelligent Systems at Huazhong University of Science & Technology (HUST) , supervised by Prof. Guoyou Wang. I also hold a Bachelor's degree from the School of Optical and Electronic Information at HUST.

My research primarily focuses on implicit video representation, compression techniques, efficient deployment of neural networks.

I proposed an image-wise implicit neural representation for videos and have been building an implicit framework based on it. I envision this implicit space functioning similarly to the Fourier Transform in signal processing, offering new perspectives and facilitating various tasks in video processing. These tasks include compression, enhancement, processing, analysis, and generation of videos. Additionally, this framework can be applied to many other types of sequential data, such as aerial and medical imagery videos, dynamic point clouds.

Projects by areas:    Neural Representation     Efficent Architecture     Object Detection    

Fast Encoding and Decoding for Implicit Video Representation
ECCV 2024 [project page] [preprint paper] [code]
Hao Chen, Saining Xie, Ser-Nam Lim, Abhinav Shrivastava

We propose NeRV-Enc, which encodes videos 104 times faster than its predecessor NeRV, utilizing hyper-networks. Additionally, we introduce NeRV-Dec, which decodes video 8.9 times faster than NeRV via parallel decoding, and is 11 times faster compared to the H.264 codec.

HNeRV: A Hybrid Neural Representation for Videos
CVPR 2023 [project page] [paper] [code]
Hao Chen, Matt Gwilliam, Ser-Nam Lim, Abhinav Shrivastava

We propose a hybrid video neural representation and a evenly distributed neural network to improve modeling capacity and introduce internal generalization.

Towards Scalable Neural Representation for Diverse Videos
CVPR 2023 [project page] [paper] [code]
Bo He, Xitong Yang, Hanyu Wang, Zuxuan Wu, Hao Chen, Shuaiyi Huang, Yixuan Ren, Ser-Nam Lim, Abhinav Shrivastava

We propose D-NeRV, a novel neural representation framework designed to encode large-scale and diverse videos.

CNeRV: Content-adaptive Neural Representation for Visual Data
BMVC 2022 (Oral) [project page] [paper]
Hao Chen, Matt Gwilliam, Bo He, Ser-Nam Lim, Abhinav Shrivastava

We propose a hybrid video neural representation with content-adaptive embedding to introduce internal generalization.

NeRV: Neural Representations for Videos
NeurIPS 2021 [project page] [paper] [code]
Hao Chen, Bo He, Hanyu Wang, Yixuan Ren, Ser-Nam Lim, Abhinav Shrivastava

We propose an image-wise neural representation for videos, which achieves good compression results and fast decoding speed.

Acknowledge

I appreciate everyone who helped me or encouraged me throughout my life, especially Prof. Abhinav Shrivastava, Prof. Guoyou Wang, Prof. Yu Qiao, Prof. Yali Wang, Prof. Xiang Bai and all good friends I met at China and the US.


The website template was borrowed from Ben Mildenhall.