About Me

I am a second year PhD student at Tandon School of Engineering of New York University, working with Prof. Yi Fang. Before this, I got my Master’s degree in Data Science and Information Technology at TBSI, Tsinghua University, supervised by Prof. Lu Fang. And I got my Bachelor’s degree at Computer School of Wuhan University.

My research interests are robotics, 3D vision and world model, especially:

Intelligent embodied robot navigation with various goals.
Leverage 3D information (point cloud, gaussian splatting) to help robots have better performance.
Robot foundation world modeling.

News

[August. 2025] My paper TransGI is accepted by IEEE TVCG!

Publications

	Hierarchical Scoring with 3D Gaussian Splatting for Instance Image-Goal Navigation Yijie Deng, Shuaihang Yuan**, Geeta Chandra Raju Bethala, Anthony Tzes, Yu-Shen Liu (ArXiv 2025) [ArXiv]
	MapBERT: Bitwise Masked Modeling for Real-Time Semantic Mapping Generation Yijie Deng, Shuaihang Yuan**, Congcong Wen, Hao Huang, Anthony Tzes, Geeta Chandra Raju Bethala, Yi Fang (ArXiv 2025) [ArXiv]
	TransGI: Real-Time Dynamic Global Illumination With Object-Centric Neural Transfer Model Yijie Deng, Lei Han, Lu Fang (IEEE TVCG) [Paper] [Arxiv]
	RealLiFe: Real-Time Light Field Reconstruction via Hierarchical Sparse Gradient Descent Yijie Deng, Lei Han**, Tianpeng Lin, Lin Li, Jinzhi Zhang, Lu Fang (ArXiv 2025) [ArXiv]

Research Experience

January 2025 - June 2025

Developed a gaussian splatting-based indoor instance image goal navigation method that can accurately locate the target image in a big indoor scene.
Build the appearance and feature gaussian splatting of a scene using a self-explored agent. And then use a hierarchical scoring method to locate the target image in both global semantic-level and local geometric-level.
Achieve the sota instance image goal navigation success rate compared to baseline methods, and is applied to a humanoid robot to prove its real-world application on real robots.

September 2024 - December 2024

Developed real-time indoor map generation method for efficient object-goal navigation.
Leverage the bit-wise autoencoder to learn the map patterns of the BEV map of indoor scenes and a masked transformer to imagine the unexplored indoor map to discover the targets in advance.
Achieve real-time map generation speed and the sota object navigation success rate.

Real-Time Cloud Rendering

April 2023 - March 2024

Developed a novel rendering system, designed specifically for high-quality, real-time graphics on low-end devices using cloud-based powerful GPUs.
Innovated in compression techniques and rendering effect decomposition, creating a compact representation for object-centric radiance transfer fields. This approach enables efficient, user-independent global illumination with low-frequency detail, alongside shared specular importance sampling for dynamic, high-frequency reflections.
Utilized Mitsuba3 rendering engine for demonstrating capabilities in Python, achieving 5 FPS at 1920x1080 resolution, maintaining quality comparable to ray tracing. Currently implementing a real-time demo using NVIDIA Falcor and CUDA for enhanced performance and efficiency.

Real-Time Light Field Reconstruction via Hierarchical SGD

October 2022 - April 2023

Develop an algorithm for real-time light field video generation supporting 3D display and VR devices.
Employed sparse light field gradients to boost performance for a balanced trade-off between quality and efficiency.
Achieved by far the fastest light field generation methods 400x faster than the baseline DeepView with a 2 dB higher PSNR compared to baseline novel view synthesis methods (IBRNet, MVSNeRF, ENeRF).

Education

PhD., Computer Science and Robotics
September 2024 - Present, New York University, US.
M.Phil., Data Science and Information Technology
September 2021 - June 2024, Tsinghua University, China.
B.Eng., Computer Science
September 2017 - June 2021, Wuha University, China.

Honors & Awards

2021, Outstanding Graduate in School of Computer Science
2019, National English Competition for College Students, Second Prize of Hubei Province
2017-2021, National Encouragement Scholarship
2017-2021, First-class Scholarship for Excellent Students
2018, National Mathematics Competition for College Students, Second Prize of Hubei Province
2018, Nationally Televised CCTV New Hope English Speech Competition, First Prize of Hubei Province

	Hierarchical Scoring with 3D Gaussian Splatting for Instance Image-Goal Navigation Yijie Deng, Shuaihang Yuan**, Geeta Chandra Raju Bethala, Anthony Tzes, Yu-Shen Liu (ArXiv 2025) [ArXiv]
	MapBERT: Bitwise Masked Modeling for Real-Time Semantic Mapping Generation Yijie Deng, Shuaihang Yuan**, Congcong Wen, Hao Huang, Anthony Tzes, Geeta Chandra Raju Bethala, Yi Fang (ArXiv 2025) [ArXiv]
	TransGI: Real-Time Dynamic Global Illumination With Object-Centric Neural Transfer Model Yijie Deng, Lei Han, Lu Fang (IEEE TVCG) [Paper] [Arxiv]
	RealLiFe: Real-Time Light Field Reconstruction via Hierarchical Sparse Gradient Descent Yijie Deng, Lei Han**, Tianpeng Lin, Lin Li, Jinzhi Zhang, Lu Fang (ArXiv 2025) [ArXiv]

Yijie Deng