Qi Zhang (张琦)
I am currently a Lead Researcher in the Imaging Algorithm Center of Vivo. Our group, as the core algorithm team, utilizes 3D and AIGC-based technologies to enhance the photography quality and user experience of smartphones. I do research in 3D computer vision and AI generated content (AIGC), where my interests focus on neural rendering, 3D Gaussian Splatting (3DGS), 3D reconstruction and modeling, and visual diffusion models.
During 2021–2024, I was a Researcher in Tencent AI Lab from Jun. 2021. Before that, I received my Ph.D. degree from the School of Computer Science of Northwestern Polytechnical University in 2021, supervised by Prof. Qing Wang. I was a visiting student at the Australian National University (ANU) between Jul. 2019 and Aug. 2020, supervised by Prof. Hongdong Li. I received the Outstanding Doctoral Dissertation Award Nominee from China Computer Federation (CCF) in 2021. I also won the 2021 ACM Xi'an Doctoral Dissertation Award and the 2023 NWPU Doctoral Dissertation Award.
Vivo (Xi'an & Hangzhou) is hiring multiple researchers and interns for projects on 3D reconstruction (e.g. NeRF, 3DGS, and NVS) and AIGC-based generation (including image or video diffusion models). Feel free to contact me via e-mail and WeChat.
News
- 2025.04 Serving as an Area Chair for NeurIPS 2025!
- 2025.03 3 papers accepted to CVPR 2025!
- 2024.08 LTM-NeRF accepted to TPAMI!
- 2024.07 4 papers accepted to ECCV 2024, including one oral paper!
- 2024.03 5 papers accepted to CVPR 2024!
- 2024.02 DINER extended to TPAMI!
- 2023.12 NeIF accepted to AAAI 2023
- 2023.08 LoD-NeuS accepted to SIGGRAPH Asia 2023
- 2023.07 Pyramid NeRF accepted to IJCV 2023
- 2023.03 7 papers (with 1 highlight) accepted to CVPR 2023!
- 2022.08 1 paper (journal track) accepted to SIGGRAPH Asia 2022
- 2022.03 4 papers accepted to CVPR 2022
- 2021.12 CCF Outstanding Doctoral Dissertation Award Nominee
Tech Transfer
Research
Neural Rendering
NeRF, light fields, and physically-based inverse rendering for novel-view synthesis and relighting.
3D Gaussian Splatting
Efficient 3DGS reconstruction, editing, and streaming for real-time radiance-field rendering.
AIGC & Diffusion
Image/video diffusion priors for super-resolution, generation, and reference-guided 3D refinement.
Computational Photography
Light field imaging and camera modeling, image rectification, super-resolution, and learning-based photography pipelines.
Publications
Below is a complete list of my publications. Papers marked with a Representative paper badge are my representative works. The IEEE TPAMI and IJCV are top journals in the field of computer vision and computational photography. CVPR is the premier conference in the computer vision research community.
HiCo-GS: Hierarchical Context Aggregation and Geometric Consistency for Octree Gaussian Splatting
Octree-based 3DGS with hierarchical context and geometric consistency for efficient, high-fidelity scene reconstruction.
3D Gaussian SplattingACM Multimedia (ACMMM), 2026
Stream4D: Streaming 4D Reconstruction with Training-free Motion Decoupling via Causal KV-Cache Attention
Training-free streaming 4D reconstruction that decouples motion via causal KV-cache attention for online dynamic scenes.
3D Gaussian SplattingACM Multimedia (ACMMM), 2026
GSFixer: Improving 3D Gaussian Splatting with Reference-Guided Video Diffusion Priors
Refines 3D Gaussian Splatting using reference-guided video diffusion priors to recover missing or degraded detail.
3D Gaussian SplattingAIGC & DiffusionInternational Conference on Machine Learning (ICML), 2026
OnlinePG: Online Open-Vocabulary Panoptic Mapping with 3D Gaussian Splatting
Online open-vocabulary panoptic mapping built on 3D Gaussian Splatting for real-time semantic scene understanding.
3D Gaussian SplattingCVPR, 2026
TinySR: Pruning Diffusion for Real-World Image Super-Resolution
Accelerates real-world image super-resolution by pruning diffusion models to a single efficient step.
Computational PhotographyAIGC & DiffusionCVPR, 2026
LiveMoments: Reselected Key Photo Restoration in Live Photos via Reference-guided Diffusion
Restores reselected key photos in Live Photos via reference-guided diffusion for high-quality moment capture.
Computational PhotographyAIGC & DiffusionInternational Conference on Learning Representations (ICLR), 2026
Finer++: Building a Family of Variable-periodic Functions for Activating Implicit Neural Representation
A family of variable-periodic activation functions that tunably control spectral bias in implicit neural representations.
Neural RenderingIEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2026
Rhino: Regularizing the Hash-based Implicit Neural Representation
Regularizes hash-based implicit neural representations to improve robustness and reconstruction stability.
Neural RenderingScience China Information Sciences 69 (1), 112101, 2026
One-shot Refiner: Boosting Feed-forward Novel View Synthesis via One-step Diffusion
Boosts feed-forward novel-view synthesis with a one-step diffusion refinement for sharper, consistent views.
3D Gaussian SplattingAAAI 2026
GUS-IR: Gaussian Splatting with Unified Shading for Inverse Rendering
Unified-shading Gaussian Splatting that jointly models geometry, material and lighting for inverse rendering.
3D Gaussian SplattingNeural RenderingIEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2025
Mitigating Ambiguities in 3D Classification with Gaussian Splatting
Resolves ambiguous 3D classification by leveraging Gaussian Splatting as a structured 3D prior.
3D Gaussian SplattingCVPR, 2025
Mani-GS: Gaussian Splatting Manipulation with Triangular Mesh
Manipulates 3D Gaussian Splatting scenes through editable triangular meshes for intuitive object control.
3D Gaussian SplattingCVPR, 2025
TSD-SR: One-Step Diffusion with Target Score Distillation for Real-World Image Super-Resolution
One-step real-world super-resolution via target score distillation from a pretrained diffusion teacher.
Computational PhotographyAIGC & DiffusionCVPR, 2025
H2O-NeRF: Radiance Fields Reconstruction for Two-Hand-Held Objects
Reconstructs radiance fields of two-hand-held objects with a dedicated two-hand interaction formulation.
Neural RenderingIEEE Transactions on Visualization and Computer Graphics (TVCG), 2025
BokehDiff: Neural Lens Blur with One-Step Diffusion
Produces lens blur (bokeh) from a single image with a one-step diffusion approach to depth-of-field control.
AIGC & DiffusionIEEE/CVF International Conference on Computer Vision (ICCV), 9508-9518, 2025
IR-Pro: Baking Probes to Model Indirect Illumination for Inverse Rendering of Scenes
Bakes probe volumes to efficiently model indirect illumination for scene-level inverse rendering.
Neural RenderingIEEE Signal Processing Letters, 2025
LTM-NeRF: Embedding 3D Local Tone Mapping in HDR Neural Radiance Field
Embeds 3D local tone mapping inside HDR neural radiance fields for view-dependent high-dynamic-range look.
Neural RenderingIEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2024
Analytic-Splatting: Anti-Aliased 3D Gaussian Splatting via Analytic Integration
Anti-aliased 3D Gaussian Splatting via analytic integration, removing splat-level aliasing artifacts.
3D Gaussian SplattingECCV, 2024
Neural Poisson Solver: A Universal and Continuous Framework for Natural Signal Blending
A universal continuous framework for natural signal blending using a neural Poisson solver.
Neural RenderingECCV, 2024
Physically Plausible Color Correction for Neural Radiance Fields
Performs physically plausible color correction of neural radiance fields to fix view-dependent tint.
Neural RenderingECCV, 2024
Head360: Learning a Parametric 3D Full-Head for Free-View Synthesis in 360°
Learns a parametric 3D full-head model enabling free-view synthesis in 360 degrees from limited input.
Neural RenderingECCV, 2024
Advances in 3D Generation: A Survey
A comprehensive survey summarizing recent advances, taxonomies and open problems in 3D generation.
AIGC & DiffusionarXiv, 2024
UV Gaussians: Joint Learning of Mesh Deformation and Gaussian Textures for Human Avatar Modeling
Jointly learns mesh deformation and Gaussian textures for drivable, editable human avatar motion.
3D Gaussian SplattingKnowledge-Based Systems, 2025
GS-IR: 3D Gaussian Splatting for Inverse Rendering
Extends 3D Gaussian Splatting to inverse rendering, decomposing scene appearance into physical components.
3D Gaussian SplattingNeural RenderingCVPR, 2024
HumanNorm: Learning Normal Diffusion Model for High-quality and Realistic 3D Human Generation
Trains a normal diffusion model that lifts 2D human images to high-quality, realistic 3D human geometry.
AIGC & DiffusionCVPR, 2024
FINER: Flexible Spectral-bias Tuning in Implicit Neural Representation by Variable-periodic Activation Functions
Flexibly tunes spectral bias in implicit neural representation through variable-periodic activations.
Neural RenderingCVPR, 2024
HumanRef: Single Image to 3D Human Generation via Reference-Guided Diffusion
Generates a 3D clothed human with realistic, view-consistent texture and geometry from a single image, guided by a reference image via Stable Diffusion.
AIGC & DiffusionCVPR, 2024
Extended HumanRef-GS to TCSVT, 2025
ConTex-Human: Free-View Rendering of Human from a Single Image with Texture-Consistent Synthesis
Renders free-view humans from one image via texture-consistent synthesis for coherent novel views.
Neural RenderingCVPR, 2024
NeIF: A Pre-convolved Representation for Plug-and-Play Neural Illumination Fields
A pre-convolved plug-and-play neural illumination field representation for efficient relighting.
Neural RenderingAAAI, 2024
Disorder-invariant Implicit Neural Representation
Makes implicit neural representations invariant to sample disorder for stable, permutation-robust fitting.
Neural RenderingIEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2024
Anti-Aliased Neural Implicit Surfaces with Encoding Level of Detail
Builds anti-aliased neural implicit surfaces with an encoding level of detail for clean geometry.
Neural RenderingSIGGRAPH Asia, 2023
Pyramid NeRF: Frequency Guided Fast Radiance Field Optimization
Accelerates radiance-field optimization using frequency-guided, pyramid-level fast convergence.
Neural RenderingInternational Journal of Computer Vision (IJCV), 2023
Wide-angle Rectification via Content-aware Conformal Mapping
Corrects wide-angle distortion via content-aware conformal mapping that preserves scene structure.
Computational PhotographyCVPR, 2023
Inverting the Imaging Process by Learning an Implicit Camera Model
Learns an implicit camera model that inverts the imaging process for calibration-free correction.
Neural RenderingCVPR, 2023
Local Implicit Ray Function for Generalizable Radiance Field Representation
A generalizable radiance-field representation using local implicit ray functions for few-shot views.
Neural RenderingCVPR, 2023
DINER: Disorder-Invariant Implicit Neural Representation
A disorder-invariant implicit neural representation robust to unstructured, permuted training samples.
Neural RenderingCVPR, 2023
Local-to-Global Registration for Bundle-Adjusting Neural Radiance Fields
Bundle-adjusts neural radiance fields via local-to-global registration for consistent multi-view fusion.
Neural RenderingCVPR, 2023
UV Volumes for Real-time Rendering of Editable Free-view Human Performance
Real-time editable free-view human performance rendering using UV volume primitives.
Neural RenderingCVPR, 2023
Fine-Grained Face Swapping via Regional GAN Inversion
Swaps faces with fine-grained regional GAN inversion for localized, identity-preserving editing.
AIGC & DiffusionCVPR, 2023
Neural Parameterization for Dynamic Human Head Editing
Enables dynamic human head editing through a learnable neural parameterization.
Neural RenderingACM Transactions on Graphics, 2022
Stereo Unstructured Magnification: Multiple Homography Image for View Synthesis
Synthesizes novel views from stereo via multiple-homography unstructured magnification.
Neural RenderingarXiv, 2022
HDR-NeRF: High Dynamic Range Neural Radiance Fields
Reconstructs high-dynamic-range neural radiance fields from multi-exposure images.
Neural RenderingCVPR, 2022
Hallucinated Neural Radiance Fields in the Wild
Hallucinates complete neural radiance fields of in-the-wild scenes from sparse or casual captures.
Neural RenderingCVPR, 2022
Deblur-NeRF: Neural Radiance Fields from Blurry Images
Reconstructs neural radiance fields directly from blurry images by modeling the blur process.
Neural RenderingCVPR, 2022
FENeRF: Face Editing in Neural Radiance Fields
Edits faces inside neural radiance fields with semantic control over identity and appearance.
Neural RenderingCVPR, 2022
Ray-Space Epipolar Geometry for Light Field Cameras
Derives ray-space epipolar geometry tailored to light field cameras for view analysis.
Computational PhotographyIEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2022
3D Scene Reconstruction with an Un-calibrated Light Field Camera
Reconstructs 3D scenes from an un-calibrated light field camera without prior calibration.
Computational PhotographyInternational Journal of Computer Vision (IJCV), 2021
Full View Optical Flow Estimation Leveraged from Light Field Superpixel
Estimates full-view optical flow leveraging light field superpixels for dense correspondence.
Computational PhotographyIEEE Transactions on Computational Imaging, 2019
Ray-Space Projection Model for Light Field Camera
A ray-space projection model specifically for light field camera geometry.
Computational PhotographyCVPR, 2019
4D Light Field Superpixel and Segmentation
Superpixel and segmentation for 4D light fields enabling efficient structural decomposition.
Computational PhotographyIEEE Transactions on Image Processing (TIP), 2019
CVPR, 2017
A Generic Multi-Projection-Center Model and Calibration Method for Light Field Camera
A generic multi-projection-center model with calibration for light field cameras.
Computational PhotographyIEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2019