Shuntian Zheng
About me
I am a second-year PhD student in the Department of Computer Science at the 果冻传媒, supervised by Dr Yu GuanLink opens in a new window and partially funded by the Department.
Research and Technical Expertise
My current research focuses on privacy-preserving human perception from wireless signals, including WiFi and mmWave sensing. In particular, I study physics-guided representations and preprocessing methods for robust 3D human pose estimation.My broader AI experience includes:
- Multimodal and video understanding
Before starting my PhD, I worked on computational photography at ZTE in Nanjing and Xiaomi in Shanghai, with a focus on automatic photo exposure and clarity enhancement. This experience shaped my interest in translating research ideas into practical, efficient systems.
I always hope to design a unique processing solution based on the physical properties of the special research object, to achieve higher usability and lower complexity, and to realize the explainability of AI methods.
I have open-sourced my millimeter-wave human sensing solution: https://github.com/DandongExpress/.
I have also open-sourced several other projects to help beginners in this field quickly get started with this task, become familiar with baseline comparisons and optimizations, and build their own better solutions.
In my spare time, I like to read history.
Education
- The 果冻传媒, Coventry, UK
PhD student, 3D Human Pose Estimation via mmWave Sensing (Oct 2024 - Now)
- Beijing University of Posts and Telecommunications, Beijing, China
BEng, Computer Science_ (Sep 2020 - Jun 2024)
Publication
[PAPER] Rethinking PPG-based Sleep Staging: Datasets, Metrics, and Benchmarks(arXiv: 2608.00943)
Shuntian Zheng, Jiawei Wang, Cong Fu, Huan Yu, Chen Chen, Yu Guan, Sai Gu
[PAPER] Doppler Prompting for Stable mmWave-based Human Pose Estimation Preprocessing(accepted by ICML 2026, Code Open-sourced)
Shuntian Zheng, Jiaqi Li, Xiaoman Lu, He Shuai, Yu Guan
[PAPER] Why Learn What Physics Already Knows? Realizing Agile mmWave-based Human Pose Estimation via Physics-Guided Preprocessing (accepted by ICME 2026, Code Open-sourced)
Shuntian Zheng, Jiaqi Li, Minzhe Ni, Xiaoman Lu, Yu Guan
[PAPER] Person Parametric Physics-informed Representation for mmWave-based Human Pose Estimation (accepted by UbiComp/IMWUT 2026, Code Open-sourced)
Shuntian Zheng, Jiaqi Li, Guangming Wang, Minzhe Ni, Arnad Palit, Giovanni Montana, Yu Guan
[PAPER] Keeping the Evidence Chain: Semantic Evidence Allocation for Training-Free Token Pruning in Video Temporal Grounding(accepted by ECCV 2026)
Jiaqi Li, Shuntian Zheng, Yixian Shen, Jia-Hong Huang, Xiaoman Lu, Minzhe Ni, Yu Guan
[PAPER] Towards Mitigating Modality Bias in Vision-Language Models for Temporal Action Localization (accepted by ACL 2026)
Jiaqi Li, Guangming Wang, Shuntian Zheng, Minzhe Ni, Xiaoman Lu, Guanghui Ye, Yu Guan
[PAPER] DA3Attacker: A Diffusion-based Attacker against Aesthetics-oriented Black-box Models (accepted by TIP 2025)
Shuai He, Shuntian Zheng, Anlong Ming, Yanni Wang, Huadong Ma
[PAPER] lEA: Pixel-level Image Exposure Assessment Using Zone Exposure Theory (accepted by NIPS 2024)
Shuai He, Shuntian Zheng, Anlong Ming, Banyu Wu, Huadong Ma
[PAPER] EAT: An Enhancer for Aesthetics-Oriented Transformers (accepted by ACM MM 2023 as oral)
Shuai He, Shuntian Zheng, Anlong Ming, Haobin Zhong, Huadong Ma
[PAPER] Delegate Transformer for Image Color Aesthetics Assessment (accepted by ICCV 2023)
Shuai He, Anlong Ming, Yaqi Li, Shuntian Zheng, Jinyuan Sun, Huadong Ma
[PAPER] Controlled Node Dialogue in IoT Networks Based on Nonlocal Orthogonal Product States (accepted by JOCN 2022 )
Bichen Che, Shuntian Zheng, Zhao Dou, Xiubo Chen, Lixiang Li, Jian Li, Yixian Yang
[STANDARD] Metadata for Image Aesthetics Assessment with Aesthetic Attributes in Mobile Terminal Computational Photography System (consented by the International Telecommunication Union)
[PATENT] Adversarial attack methods, devices, and storage media for image aesthetic evaluation models
Main Inventor
[PATENT] Image Aesthetic Evaluation Methods, Devices, Electronic Equipment, and Storage Media
Main Inventor
[PATENT] Decentralized Management Technology for University Scholarships Based on Digital Currency
Main Inventor
Open Source
(1) RadarHPE-Toolbox
A unified, physics-guided library for millimetre-wave radar Human Pose Estimation.
Learn the signal. Train the model. Benchmark fairly. Contribute your method.
It provides FMCW fundamentals, shared dataset loaders (HuPR / XRF55 / mmRadPose / MMVR), standard metrics (MPJPE, PA-MPJPE, MPJVE, AKV), and a registry-based model zoo including methods from our ICML / ICME / UbiComp papers.
鈫� https://github.com/DandongExpress/RadarHPE-Toolbox
(2) VideoRadarSim
A companion simulator that synthesises trainable mmWave radar data from video or 3D human meshes.
From ordinary video to privacy-preserving radar signals鈥攑hysics-first, toolbox-ready.
It targets the data bottleneck in wireless sensing: expensive radar collection is replaced by a physics-guided pipeline (mesh 鈫� scatterers 鈫� FMCW IF 鈫� Range鈥揂ngle鈥揇oppler heatmaps) with export formats designed for RadarHPE-Toolbox.
鈫� https://github.com/DandongExpress/VideoRadarSim
Individual paper code releases (PPPR, PULSE, Agile-MmWave-HPE, etc.) are linked from the corresponding publications and from the Toolbox model zoo.
Teaching
Guest Lecturer, CS345 Sensor Networks and Mobile Data Communications (2025/26)
Department of Computer Science, 果冻传媒
Invited to deliver guest lectures and teaching sessions on wireless sensing and related topics for undergraduate students.
Visit
Hosted by Dr Chen Chen. Research focus: non-contact sleep-stage detection and human-health sensing.
Professional Service
Competition
Contact
Email:
Room: MB4.20