Robust AI in the Wild Lab · HKUST(GZ) · AI Thrust

RAW Lab
@ HKUST-GZ

We sit at the intersection of physics-based vision, computational photography, and physical AI — building systems that perceive, reconstruct, and interact with the physical world grounded in the principles of light and imaging.

RAW is three things at once: the unprocessed signal straight from the camera sensor, Robust AI in the Wild, and — in the spirit of our research — 本质, the Chinese word for essence. We study the world as it truly is, before any approximation is made.

Our Research Publications
Scroll
Who We Are

Physics-grounded vision
for the real world

The RAW Lab (Robust AI in the Wild) is a research group at the Thrust of Artificial Intelligence, Hong Kong University of Science and Technology (Guangzhou), led by Dr. Ziteng Cui.

We sit at the intersection of physics-based vision, computational photography, and physical AI. Rather than treating the imaging pipeline as a black box, we model light transport, sensor physics, and scene geometry as learnable, differentiable systems — enabling AI that is robust, interpretable, and deployable in the real world.

RAW SENSOR · BAYER PATTERN
Physics-Based Vision
Low-level Vision, Image Processing, Vision Robustness
3D NEURAL REPRESENTATIONS
Computational Photography
Sensor Imaging Pipeline, 3D Computer Vision
PHYSICAL AI · EMBODIED SYSTEMS
Physical AI
Hardware-in-the-loop for robotics & autonomous systems, Robotic Robustness
Research Directions

What we work on

Our work lives at the crossroads of three fields — using the physics of imaging as the bridge between visual understanding and intelligent physical interaction.

🌒

Physics-Aware 3D Neural Representations

Extending NeRF and 3DGS to handle variable illumination, participating media, and real-world degradation through physics-grounded formulations.

Radiance Field 3D Gaussian Splatting Inverse Rendering Digital Twins
📷

Camera-based Perception & RAW Vision

Building differentiable pipelines from raw sensor data to high-level tasks, bypassing human-centric ISPs to preserve physically rich photometric information.

Low-level Vision ISP Computational Photography
🤖

AI for Next-Generation Camera Design

Co-designing optics, sensors, and algorithms to produce machine-optimal visual data for robotics, autonomous driving, and embodied intelligence.

Embodied AI Camera Co-design Robustness
⚗️

Physics-Guided Generative Models

Integrating optical and geometric priors into generative AI to synthesize physically consistent scenes and enable high-quality synthetic data engines.

Simulation Data Engine World Models
🔬

Solving Inverse Problems in Imaging

Recovering intrinsic physical scene properties from images — albedo, shading, depth, material — for scientific, medical, and industrial applications.

Inverse Problems Scene Understanding Remote Sensing
📊

Physically Degraded Benchmarks

Creating evaluation frameworks that reflect real-world sensor degradation, including the RealX3D benchmark for multi-view 3D restoration and reconstruction.

RealX3D Benchmark 3D Reconstruction
Selected Publications

Recent work

T-PAMI
2026
RAW-Adapter: Adapting Pre-trained Visual Models to Camera RAW Images and A Benchmark
Ziteng Cui, Jianfei Yang, Tatsuya Harada
CVPR
2026 Finding
M3APolicy: Mutable Material Manipulation Augmentation Policy through Photometric Re-rendering
Jiayi Li, Yuxan Hu, Haoran Geng, Xiangyu Chen, Chuhao Zhou, Ziteng Cui, Jianfei Yang
CVPR
2026
MERIT: Multi-domain Efficient RAW Image Translation
Wenjun Huang, Shenghao Fu, Yian Jin, Yang Ni, Ziteng Cui, Hanning Chen, Yirui He, Yezi Liu, Sanggeon Yun, SungHeon Jeong, Ryozo Masukawa, William Youngwoo Chung, Mohsen Imani.
WACV
2026
Perception-Inspired Color Space Design for Photo White Balance Editing
Yang Cheng, Ziteng Cui#, Lin Gu, Shenghan Su, Zenghui Zhang.
CVPR
2025
Luminance-GS: Adapting 3D Gaussian Splatting to Challenging Lighting Conditions with View-Adaptive Curve Adjustment
Ziteng Cui#, Xuangeng Chu, Tatsuya Harada
NeurIPS
2025
Dr. RAW: Towards General High-Level Vision from RAW with Efficient Task Conditioning
Wenjun Huang*, Ziteng Cui*, Yinqiang Zheng, Yirui He, Tatsuya Harada, Mohsen Imani
NeurIPS
2025
I2-NeRF: Learning Neural Radiance Fields Under Physically-Grounded Media Interactions
Shuhong Liu, Lin Gu, Ziteng Cui, Xuangeng Chu, Tatsuya Harada
Siggraph Asia
2025
ARTalk: Speech-Driven 3D Head Animation via Autoregressive Model
Xuangeng Chu, Nabarun Goswami, Ziteng Cui, Hanqin Wang, Tatsuya Harada
ECCV
2024
RAW-Adapter: Adapting Pre-trained Visual Models to Camera RAW Images
Ziteng Cui#, Tatsuya Harada
BMVC
2024
Discovering an Image-Adaptive Coordinate System for Photography Processing
Ziteng Cui, Lin Gu, Tatsuya Harada
AAAI
2024
Aleth-NeRF: Illumination Adaptive NeRF with Concealing Field Assumption
Ziteng Cui, Lin Gu, Xiao Sun, Yu Qiao, Tatsuya Harada
ICCV
2023
Monodetr: Depth-guided transformer for monocular 3d object detection
Renrui Zhang, Han Qiu, Tai Wang, Ziyu Guo, Ziteng Cui, Yu Qiao, Hongsheng Li, Peng Gao
BMVC
2022
You Only Need 90K Parameters to Adapt Light: A Light Weight Transformer for Image Enhancement and Exposure Correction ★ Most Cited Paper of BMVC 2022
Ziteng Cui, Kunchang Li, Lin Gu, Shenghan Su, Peng Gao, Yu Qiao, Tatsuya Harada
ECCV
2022
Exploring Resolution and Degradation Clues as Self-supervised Signal for Low Quality Object Detection
Ziteng Cui, Yingying Zhu, Lin Gu, Guo-Jun Qi, Xiaoxiao Li, Renrui Zhang, Zenghui Zhang, Tatsuya Harada
ICCV
2021
Multitask AET with Orthogonal Tangent Regularity for Dark Object Detection
Ziteng Cui, Guo-Jun Qi, Lin Gu, Shaodi You, Zenghui Zhang, Tatsuya Harada
Full publication list →
Team

People

Photo
Ziteng Cui
Principal Investigator · Assistant Professor

Dr. Ziteng Cui is an incoming Assistant Professor at the AI Thrust, HKUST(GZ). He received his Ph.D. from the University of Tokyo (2025) under Prof. Tatsuya Harada, where he was awarded the Dean's Award of the Graduate School of Engineering. His research sits at the intersection of physics-based vision, computational photography, and physical AI. He like pixel, vision and light, publish papers at CVPR, ECCV, NeurIPS, ICCV, AAAI, T-PAMI, IJCV, Siggraph Asia, etc.

Homepage Google Scholar GitHub

We are actively looking for motivated PhD students, Mphils and research interns
interested in computational photography, low-level vision, 3D vision, and physical AI.

Get in Touch
Contact

Find us

RAW Lab is located at the Thrust of Artificial Intelligence, Hong Kong University of Science and Technology (Guangzhou) — one of the fastest-growing research universities in the Greater Bay Area.

We welcome collaborations from academia and industry. If you are interested in our work or are a prospective student, please reach out by email.

✉️
Email
cui@mi.t.u-tokyo.ac.jp
📍
Address
Thrust of Artificial Intelligence
HKUST(Guangzhou), Nansha, Guangzhou 511458, China
🔗
Web
cuiziteng.github.io/raw-lab