What If? Guided Diffusion as a Tool for Scientific Inquiry into Tropical Cyclones
In preparation
Steer an ensemble weather model toward a target storm track or size, to explore rare but high‑impact scenarios
On the job market for full-time roles from 2027. CV
Teaching generative models to respect light, geometry, and the camera.
CS PhD student at the University of Maryland, advised by Christopher Metzler.
I make image and video generation controllable (light, focus, reflections, camera, and more) and keep it faithful to the real scene by grounding it in physics and geometry. Earlier work covers restoration, image quality, and robust vision–language models.
Previously at X-Pixel with Chao Dong and Jinjin Gu; BEng from CUHK-Shenzhen.
Selected work
Four projects from my papers. Move your cursor across the first two to move the light and pull focus.
News
-
Our work on vision–language robustness under adverse imaging was accepted to CoLM 2026. Congratulations to Tianfu and Mingyang.
-
Large-Scale Light Field Synthesis from Videos was accepted to ECCV 2026.
-
Back at the Adobe NextCam team for a second summer internship.
-
Parametric Shadow Control was accepted to ICCV 2025.
-
Named an Outstanding Reviewer for CVPR 2025.
-
Joined the Adobe NextCam team (led by Marc Levoy) as a summer intern, mentored by Shumian Xin and Zhoutong Zhang.
-
Two papers accepted to CVPR 2025. Congratulations to Tianfu, Mingyang, and Jingxi.
Experience
-
Summer 2026
Adobe San Jose, CA
Ph.D. Research Intern, NextCam team (Marc Levoy).
Fast instruction-based local image editing with diffusion transformers: compute scales with the edited region, and everything outside it stays untouched across multi-turn edits. Paper in preparation.
-
Summer 2025
Adobe San Jose, CA
Ph.D. Research Intern, NextCam team (Marc Levoy), supervised by Shumian Xin and Zhoutong Zhang.
Turned casual videos into light fields to create physically correct bokeh training data. ECCV 2026; patent filing in process.
-
Summer 2024
Dolby Vision Lab Sunnyvale, CA
Ph.D. Research Intern, supervised by Guan-Ming Su.
Portrait lighting control for text-to-image diffusion models without light stage data. Outcome: ICCV 2025, one patent.
-
2020 – 2022
XPixel Group, SIAT Shenzhen, China
Undergraduate Research Assistant, supervised by Chao Dong.
Efficient and controllable image restoration; image quality and aesthetic assessment. Outcome: two ECCV papers, two CVPR workshops organized, one patent.
Publications Google Scholar →
* denotes equal contribution.
In preparation
2026
Enhancing Vision Language Models' Robustness Under Adverse Imaging Conditions
COLM 2026
Answer questions about foggy, rainy or dark frames directly, without restoring the image first
Real2SAM2Real: Generative 3D Caches as Complementary Context for Video Diffusion
Manuscript
Control the camera and the objects in video generation with a generative 3D cache
Large-Scale Light Field Synthesis from Videos Enables Geometrically Consistent Bokeh Editing
ECCV 2026
Render geometrically consistent bokeh on scenes that defeat other methods — transparent surfaces, thin structures, cluttered depth
2025
Parametric Shadow Control for Portrait Generation in Text-to-Image Diffusion Models
ICCV 2025
Add a shadow knob to AI portraits, without heavy compute
Flash-Split: 2D Reflection Removal with Flash Cues and Latent Separation
CVPR 2025
Separate reflection from transmission with flash cues and a diffusion prior
Repurposing Pre-trained Video Diffusion Models for Event-based Video Interpolation
CVPR 2025
Turn event streams into smooth video with a video diffusion model
2024
Temporally Consistent Atmospheric Turbulence Mitigation with Neural Representations
NeurIPS 2024
Low-pass the temporal dimension to strip atmospheric turbulence out of video
Flash-Splat: 3D Reflection Removal with Flash Cues and Gaussian Splats
ECCV 2024
Separate reflections in 3D using flash-induced cues
CodedEvents: Optimal Point-Spread-Function Engineering for 3D-Tracking with Event Cameras
CVPR 2024
Engineer the PSF so event cameras encode depth for free
2023
Assessor360: Multi-sequence Network for Blind Omnidirectional Image Quality Assessment
NeurIPS 2023
Assess 360° image quality the way a viewer actually explores the sphere
Snow Removal in Video: A New Dataset and A Novel Method
ICCV 2023
Remove snow from video consistently across frames, with a new dataset of realistic snow rendered onto real footage
2022
Super-resolution by predicting offsets: An ultra-efficient super-resolution network for rasterized images
ECCV 2022
Super-resolve by predicting where to sample, not what to generate
Efficient image super-resolution using vast-receptive-field attention
AIM Challenge @ ECCV 2022
Widen the receptive field with attention instead of depth
Blueprint separable residual network for efficient image super-resolution
NTIRE Challenge @ CVPR 2022 NTIRE '22 Winner
Super-resolve on a tight compute budget with blueprint separable convolutions
2021
Toward interactive modulation for photo-realistic image restoration
CVPRW 2021
One GAN for a range of blur and noise levels, with interactive control
2020
PIPAL: A Large-Scale Image Quality Assessment Dataset for Perceptual Image Restoration
ECCV 2020
Score image quality on the artifacts that modern restoration models actually produce
PaperDatasetNTIRE'21 ChallengeNTIRE'22 Full-RefNTIRE'22 No-Ref
Chief vibe officer
Presides over the Maryland headquarters. Contributes to every paper by lying on the keyboard. Has strong feelings about not being in the acknowledgements.
