Jingchuan Hu

Hi, I am Jingchuan Hu, you can call me Eric.

I'm now working at Alibaba as an engineer. I obtained my Master's degree from SJTU, during which time I conducted research at the MediaLab under the supvervision of Prof. Li Song. Before that, I received My B.Sc.degree in the same university.

My research interests include FreeView Video, Neural Rendering, Digital Human and 3D Vision. I am also exploring the power of AIGC and is an enthusiast of Stable Diffusion. Please feel free to contact me by email!

Blogs  /  Email  /  Github  /  CV  / 

My current research interests lie in the intersection of 3D vision and Graphics, specifically in FreeView Video, Neural Rendering, Digital Human and 3D Vision, etc.

In this paper, we introduce a real-time FVV synthesis system based on DIBR and a depth estimation network. Specifically, our system includes a multi-view synchronous video capture system, a multi-stage depth estimation network, a GPU-accelerated DIBR algorithm, and a virtual view parameters generation method. We also introduce a high-quality multi-view RGB-D synchronous dataset for the training of our depth estimation network. As far as we know, our system provides the first complete real-time FVV solution at a fixed field based on DIBR and a depth estimation network.

In this paper, we novelly introduce a CNN-based view interpolation algorithm to synthesis dense virtual views in real time. Based on this, we also build an end-to-end live free-viewpoint system with a multi-user oriented streaming strategy.

In this paper, a low-cost, real-time system: LiveRecon3D, is presented, with multiple RGB-D cameras connected to one single computer. The goal of the system is to provide an interactive frame rate for 3D content capture and rendering at a reduced cost.

In this paper, we present a new dynamic RGB-D video dataset with up to 12 views. Our dataset consists of 13 groups of dynamic video sequences, taken in the same scene.

In this paper, an open-solution, low-latency, real-time 3D reconstruction system is presented. The proposed system runs in a simple architecture where multiple Kinect v2 sensors are connected to a single computer. Experimental results are presented to verify the system's effectiveness concerning the 3D reconstruction visual quality and real-time performance.

We reconstructed and rendered animated human characters by neural rendering, using SMPL as a priori. Specifically, we decoupled the process into geometric and texture reconstruction. We implicitly represented the geometry of the human body using the signed distance function (SDF), while we learned fine textures through appearance blending. The animated human characters we finally reconstructed achieved good results in both geometry and texture.

We Shoot multi-view videos in real time with a sparsely arranged multi-view camera system. The captured highlights are converted into bullet time video in real time by a free view synthesis algorithm. Bullet time video allows the audience to have the same immersive experience as in the movie "The Matrix".

We used a rather traditional Depth-Image Based Rendering algorithm, which performs interpolation by warping neighboring view RGB values to novel view based on depth values, and then blend them together. The system is optimized and can run in real-time.

MS. in Information Engineering, EE Department   2020.09 - 2023.03

B.Eng. in Information Engineering, EE Department   2016.09 - 2020.06

Applied Algorithm Engineering   2023.03 - Present

3D Vision Algorithm Engineering Intern   2022.06 - 2023.02

I have a passion for fitness and enjoy challenging myself with various workout routines. My focus is primarily on weight training, specifically barbell exercises such as bench press, squats, and deadlifts. My current personal records include a 90kg bench press, a 120kg squat, and a 120kg deadlift. Fitness is not just a hobby for me, it is a way of life and I am excited to see what new challenges and accomplishments lie ahead. I am excited to continue this journey of growth and self-improvement, both in and out of the gym.

Website template borrowed from Jon Barron's personal page Here