- Algebraic Geometry and Number Theory
- Geometry and complex manifolds
- Geometric Analysis and Curvature Flows
- Advanced Algebra and Geometry
- Image Enhancement Techniques
- Image and Signal Denoising Methods
- Homotopy and Cohomology in Algebraic Topology
- Advanced Image Processing Techniques
- Color Science and Applications
- Generative Adversarial Networks and Image Synthesis
- Recommender Systems and Techniques
- Computer Graphics and Visualization Techniques
- Advanced Vision and Imaging
- Digital Media Forensic Detection
- Advanced Steganography and Watermarking Techniques
- Geometric and Algebraic Topology
- Crystallography and molecular interactions
- Advanced Differential Geometry Research
- Algebraic structures and combinatorial models
- Holomorphic and Operator Theory
- Aesthetic Perception and Analysis
- Visual Attention and Saliency Detection
- Meromorphic and Entire Functions
- Commutative Algebra and Its Applications
- 3D Shape Modeling and Analysis
Centre National de la Recherche Scientifique
2021-2025
Université Côte d'Azur
2019-2025
Wenzhou Medical University
2025
Shanghai Jiao Tong University
2021-2024
Laboratoire Jean-Alexandre Dieudonné
2021-2024
Institut Universitaire de France
2024
Mayo Clinic in Arizona
2024
Laboratoire de Mathématiques
2021-2023
Shenyang Jianzhu University
2022-2023
Medical Architecture (United Kingdom)
2023
Bacteria are pivotal in the etiology of dental caries, underscoring critical importance effective plaque control mitigating tissue diseases. Superhydrophobic materials, with their exceptional properties non-wettability, antibacterial activity, and self-cleaning capabilities, present a promising approach for caries prevention. However, clinical application remains constrained by challenges achieving stable adhesion to tooth surfaces. Inspired natural mechanisms acquired salivary pellicle...
Image harmonization aims to produce visually harmonious composite images by adjusting the foreground appearance be compatible with background. When image has photographic and painterly background, task is called harmonization. There are only few works on this task, which either time-consuming or weak in generating well-harmonized results. In work, we propose a novel network consisting of dual-domain generator discriminator, harmonizes both spatial domain frequency domain. The performs using...
Let $X$ be a compact K\"ahler manifold and let $(L, \varphi)$ pseudo-effective line bundle on $X$. We first define notion of numerical dimension bundles with singular metrics, then discuss the properties this type dimension. finally prove very general Kawamata-Viehweg-Nadel vanishing theorem an arbitrary manifold.
Let <inline-formula content-type="math/mathml"> <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" alttext="upper X"> <mml:semantics> <mml:mi>X</mml:mi> <mml:annotation encoding="application/x-tex">X</mml:annotation> </mml:semantics> </mml:math> </inline-formula> be a simply connected projective manifold with nef anticanonical bundle. We prove that is product of rationally and trivial canonical As an application we describe the MRC-fibration any
Painterly image harmonization aims to insert photographic objects into paintings and obtain artistically coherent composite images. Previous methods for this task mainly rely on inference optimization or generative adversarial network, but they are either very time-consuming struggling at fine control of the foreground (e.g., texture content details). To address these issues, we propose a novel Harmonization stable Diffusion model (PHDiffusion), which includes lightweight adaptive encoder...
Image composition refers to inserting a foreground object into background image obtain composite image. In this work, we focus on generating plausible shadows for the inserted make more realistic. To supplement existing small-scale dataset, create large-scale dataset called RdSOBA with rendering techniques. Moreover, design two-stage network named DMASNet decomposed mask prediction and attentive shadow filling. Specifically, in first stage, decompose box shape prediction. second attend...
Let $X$ be a projective manifold such that the anticanonical bundle $-K_X$ is nef. We prove Albanese map $p: X \rightarrow Y$ locally isotrivial. In particular, $p$ submersion.
Abstract We establish a new extension result for twisted canonical forms defined on hypersurface with simple normal crossings of projective manifold. Some the examples presented in appendix show that bounds we obtain are sharp.
In a recent paper, W. Ou generalized the algebraic integrability criteria of Campana-P\u{a}un and Druel to compact K\"ahler setting. A key ingredient in his proof is an algebraicity criterion, extends Bost, Bogomolov-McQuillan. this note, we explain criterion.
Abstract Let X be a compact Kähler manifold such that the anticanonical bundle <m:math xmlns:m="http://www.w3.org/1998/Math/MathML"> <m:mrow> <m:mo>-</m:mo> <m:msub> <m:mi>K</m:mi> <m:mi>X</m:mi> </m:msub> </m:mrow> </m:math> $-K_{X}$ is nef. A classical conjecture claims Albanese map <m:mo>→</m:mo> <m:mi>T</m:mi> $X\to T$ submersive. We prove this if general fibre weak Fano manifold. If projective we also for fibres of dimension at most two.
In this paper, we study projective klt pairs (X, ∆) with nef anti-log canonical divisor -(K X + and their maximal rationally connected fibration ψ : Y .We prove that the numerical dimension of +∆) on coincides Xy +∆ ) a general fiber y , which is an analogue Ejiri-Gongyo's result formulated for Kodaira dimension.As corollary, obtain relation between positivity anti-canonical rational connectedness, provides sharper estimate than in Hacon-M c Kernan's question.Moreover, case being smooth,...
The goal of image harmonization is adjusting the foreground appearance in a composite to make whole harmonious. To construct paired training images, existing datasets adopt different ways adjust illumination statistics foregrounds real images produce synthetic images. However, have considerable domain gap and performances on small-scale are limited by insufficient data. In this work, we explore learnable augmentation enrich diversity for better performance. particular, our designed SYthetic...
Given a composite image, image harmonization aims to adjust the foreground illumination be consistent with background. Previous methods have explored transforming features achieve competitive performance. In this work, we show that using global information guide feature transformation could significant improvement. Besides, propose transfer foreground-background relation from real images images, which can provide intermediate supervision for transformed encoder features. Additionally,...
In this paper, we study a projective klt pair $(X, \Delta)$ with the nef anti-log canonical divisor $-(K_X+\Delta)$ and its maximally rationally connected fibration $\psi: X \dashrightarrow Y$. We prove that numerical dimension of on $X$ coincides $-(K_{X_y}+\Delta_{X_y})$ general fiber $X_y$ Y$, which is an analogue Ejiri-Gongyo's result formulated for Kodaira dimension. As corollary, reveal relation between positivity anti-canonical rational connectedness, gives sharper estimate than...
We study several questions involving relative Ricci-flat Kähler metrics for families of log Calabi-Yau manifolds. Our main result states that if p:(X,B)\to Y is a fiber space such (X_y, B|_{X_y}) generically klt, K_{X/Y}+B relatively trivial and p_*(m(K_{X/Y}+B)) Hermitian flat some suitable integer m , then p locally trivial. Motivated by in birational geometry, we investigate the regularity singular metric corresponding to family klt pairs (X_y,B_y) \kappa(K_{X_y}+B_y)=0 . Finally,...
Let $(X, \omega_X)$ be a compact K\"ahler manifold such that the anticanonical bundle $-K_X$ is nef. We prove slopes of Harder-Narasimhan filtration tangent with respect to polarization form $\omega_X^{n-1}$ are semi-positive. As an application, we give characterization rationally connected manifolds nef bundles. another simple proof surjectivity Albanese map.
Given a composite image with photographic object and painterly background, harmonization targets at stylizing the to be compatible background. Despite competitive performance of existing works, they did not fully leverage objects in artistic paintings. In this work, we explore learning from for harmonization. particular, learn mapping background style information based on With learnt mapping, can hallucinate target object, which is used harmonize encoder feature maps produce harmonized...
Painterly image harmonization aims to harmonize a photographic foreground object on the painterly background. Different from previous auto-encoder based networks, we develop progressive multi-stage network, which harmonizes composite low-level styles (e.g., color, simple texture) high-level complex texture). Our network has better interpretability and performance. Moreover, design an early-exit strategy automatically decide proper stage exit, can skip unnecessary even harmful late stages....
In this short note, we prove the Iitaka C nm -conjecture for algebraic fiber spaces over surfaces.