arXiv subjects

Fan Ji

Publications and source records attributed to Fan Ji.

9 recordsLinked to original sources

arXiv2026

LoViF 2026 The First Challenge on Unified Removal of Raindrops and Reflections: Methods and Results

This workshop paper comprehensively reviews the First Challenge on Unified Removal of Raindrops and Reflections. The challenge aims to address a frequently encountered practical problem in the field of autonomous driving, i.e., raindrop-reflection composite degradation on rainy days. This competition attracted 149 registered participants and received 12 valid final submissions with corresponding fact sheets, significantly contributing to the progress of unified removal of raindrops and reflections. All the methods are developed and evaluated on our real-shot RainDrop and ReFlection (RDRF) dataset. A detailed analysis of the submitted methods and corresponding results is provided in this report, which highlights effective approaches and provides interesting insights for future research.

cs.CV↗

arXiv2026

Self-Evolving Agentic Image Restoration via Deliberate Planning and Intuitive Execution

Real-world image restoration (IR) remains challenging due to complex and coupled degradations. While recent agentic IR frameworks leverage Large Language Models for flexible tool planning, they face two critical limitations. First, from a search scheme perspective, excessive reliance on greedy strategies fails to balance exploration and exploitation. Second, existing agentic systems underutilize information, exhibiting episodic amnesia. To address these challenges, we propose \textbf{Self-Evolving Agentic Image Restoration (SEAR)}, which formulates restoration as a sequential decision-making problem. Inspired by the dual-process theory, SEAR comprises an Intuitive Executor and a Deliberate Planner, respectively following the fast-thinking \textit{System 1} and slow-thinking \textit{System 2} principles. The Deliberate Planner employs Pruning-Aware Monte Carlo Tree Search for long-horizon reasoning, utilizing a hybrid no-reference reward and a Multimodal Large Language Model (MLLM)-based tournament to prevent metric exploitation. Complementarily, the Intuitive Executor leverages a self-evolving episodic memory indexed by degradation-aware state fingerprints. This mechanism distills expensive search trajectories into adaptive expertise, overcoming episodic amnesia while progressively amortizing cold-start exploration costs through memory reuse. Extensive experiments on synthetic and real-world benchmarks demonstrate its strong perceptual and quantitative performance.

cs.CV↗

arXiv2022

MOST-Net: A Memory Oriented Style Transfer Network for Face Sketch Synthesis

Face sketch synthesis has been widely used in multi-media entertainment and law enforcement. Despite the recent developments in deep neural networks, accurate and realistic face sketch synthesis is still a challenging task due to the diversity and complexity of human faces. Current image-to-image translation-based face sketch synthesis frequently encounters over-fitting problems when it comes to small-scale datasets. To tackle this problem, we present an end-to-end Memory Oriented Style Transfer Network (MOST-Net) for face sketch synthesis which can produce high-fidelity sketches with limited data. Specifically, an external self-supervised dynamic memory module is introduced to capture the domain alignment knowledge in the long term. In this way, our proposed model could obtain the domain-transfer ability by establishing the durable relationship between faces and corresponding sketches on the feature level. Furthermore, we design a novel Memory Refinement Loss (MR Loss) for feature alignment in the memory module, which enhances the accuracy of memory slots in an unsupervised manner. Extensive experiments on the CUFS and the CUFSF datasets show that our MOST-Net achieves state-of-the-art performance, especially in terms of the Structural Similarity Index(SSIM).

cs.CV↗

arXiv2021

PAENet: A Progressive Attention-Enhanced Network for 3D to 2D Retinal Vessel Segmentation

3D to 2D retinal vessel segmentation is a challenging problem in Optical Coherence Tomography Angiography (OCTA) images. Accurate retinal vessel segmentation is important for the diagnosis and prevention of ophthalmic diseases. However, making full use of the 3D data of OCTA volumes is a vital factor for obtaining satisfactory segmentation results. In this paper, we propose a Progressive Attention-Enhanced Network (PAENet) based on attention mechanisms to extract rich feature representation. Specifically, the framework consists of two main parts, the three-dimensional feature learning path and the two-dimensional segmentation path. In the three-dimensional feature learning path, we design a novel Adaptive Pooling Module (APM) and propose a new Quadruple Attention Module (QAM). The APM captures dependencies along the projection direction of volumes and learns a series of pooling coefficients for feature fusion, which efficiently reduces feature dimension. In addition, the QAM reweights the features by capturing four-group cross-dimension dependencies, which makes maximum use of 4D feature tensors. In the two-dimensional segmentation path, to acquire more detailed information, we propose a Feature Fusion Module (FFM) to inject 3D information into the 2D path. Meanwhile, we adopt the Polarized Self-Attention (PSA) block to model the semantic interdependencies in spatial and channel dimensions respectively. Experimentally, our extensive experiments on the OCTA-500 dataset show that our proposed algorithm achieves state-of-the-art performance compared with previous methods.

eess.IV↗

arXiv2021

Nuclear pasta and symmetry energy in the relativistic point-coupling model

Nonuniform structure of low-density nuclear matter, known as nuclear pasta, is expected to appear not only in the inner crust of neutron stars but also in core-collapse supernova explosions and neutron-star mergers. We perform fully three-dimensional calculations of inhomogeneous nuclear matter and neutron-star matter in the low-density region using the Thomas-Fermi approximation. The nuclear interaction is described in the relativistic mean-field approach with the point-coupling interaction, where the meson exchange in each channel is replaced by the contact interaction between nucleons. We investigate the influence of nuclear symmetry energy and its density dependence on pasta structures by introducing a coupling term between the isoscalar-vector and isovector-vector interactions. It is found that the properties of pasta phases in the neutron-rich matter are strongly dependent on the symmetry energy and its slope. In addition to typical shapes like droplets, rods, slabs, tubes, and bubbles, some intermediate pasta structures are also observed in cold stellar matter with a relatively large proton fraction. We find that nonspherical shapes are unlikely to be formed in neutron-star crusts, since the proton fraction obtained in $\beta$ equilibrium is rather small. The inner crust properties may lead to a visible difference in the neutron-star radius.

nucl-th↗

arXiv2021

Hadron-quark mixed phase in the quark-meson coupling model

We explore the possibility of a structured hadron-quark mixed phase forming in the interior of neutron stars. The quark-meson coupling (QMC) model, which explicitly incorporates the internal quark structure of the nucleon, is employed to describe the hadronic phase, while the quark phase is described by the same bag model as the one used in the QMC framework, so as to keep consistency between the two coexisting phases. We analyze the effect of the appearance of hadron-quark pasta phases on the neutron-star properties. We also discuss the influence of nuclear symmetry energy and the bag constant B in quark matter on the deconfinement phase transition. For the treatment of the hadron-quark mixed phase, we use the energy minimization method and compare it with the Gibbs construction. The finite-size effects like surface and Coulomb energies are taken into account in the energy minimization method; they play crucial roles in determining the pasta configuration during the hadron-quark phase transition. It is found that the finite-size effects can significantly reduce the region of the mixed phase relative to that of the Gibbs construction. Using a consistent value of B in the QMC model and quark matter, we find that hadron-quark pasta phases are formed in the interior of massive stars, but no pure quark matter can exist.

nucl-th↗

arXiv2020

Nuclear pasta in hot and dense matter and its influence on the equation of state for astrophysical simulations

We explore the properties of nuclear pasta appearing in supernova matter, i.e., matter at finite temperature with a fixed proton fraction. The pasta phases with a series of geometric shapes are studied using the compressible liquid-drop (CLD) model, where nuclear matter separates into a dense liquid phase of nucleons and a dilute gas phase of nucleons and $\alpha$ particles. The equilibrium conditions for two coexisting phases are derived by minimization of the total free energy including the surface and Coulomb contributions, which are clearly different from the Gibbs conditions for phase equilibrium due to the finite-size effects. Compared to the results considering only spherical nuclei, the inclusion of pasta phases can delay the transition to uniform matter and enlarge the region of nonuniform matter in the phase diagram. The thermodynamic quantities obtained in the present calculation with the CLD model are consistent with those in the realistic equation of state table for astrophysical simulations using the Thomas--Fermi approximation. It is found that the density ranges of various pasta shapes depend on both the temperature $T$ and the proton fraction $Y_p$. Furthermore, the nuclear symmetry energy and its density dependence may play crucial roles in determining the properties of pasta phases. Our results suggest that the pasta phase diagram is most sensitively dependent on the symmetry energy slope $L$ especially in the low-$Y_p$ and high-$T$ region.

astro-ph.HE↗

arXiv2020

Effects of symmetry energy on equation of state for simulations of core-collapse supernovae and neutron-star mergers

We construct a new equation of state (EOS) for numerical simulations of core-collapse supernovae and neutron-star mergers based on an extended relativistic mean-field model with a small symmetry energy slope $L$, which is compatible with both experimental nuclear data and recent observations of neutron stars. The new EOS table (EOS4) based on the extended TM1 (TM1e) model with $L=40$ MeV is designed in the same tabular form and compared with the commonly used Shen EOS (EOS2) based on the original TM1 model with $L=110.8$ MeV. This is convenient and useful for performing numerical simulations and examining the influences of symmetry energy and its density dependence on astrophysical phenomena. In comparison with the TM1 model used in EOS2, the TM1e model provides a similar maximum neutron-star mass but smaller radius and tidal deformability for a $1.4 M_\odot$ neutron star, which is more consistent with current constraints. By comparing the phase diagram and thermodynamic quantities between EOS4 and EOS2, it is found that the TM1e model predicts relatively larger region of nonuniform matter and softer EOS for neutron-rich matter. Significant differences between EOS4 and EOS2 are observed in the case with low proton fraction, while the properties of symmetric matter remain unchanged.

nucl-th↗

arXiv2019

Effects of nuclear symmetry energy and equation of state on neutron star properties

We study the effects of nuclear symmetry energy on the mass-radius relation and tidal deformability of neutron stars, considering the self-consistency of the equation of state (EOS). We first construct a set of unified EOSs based on relativistic mean-field models with different density dependence of the symmetry energy. For the description of pasta phases appearing in the inner crust of neutron stars, we perform a self-consistent Thomas-Fermi calculation using the same nuclear interaction as that for the uniform matter in the core. To examine possible effects from the self-consistency of the EOS on neutron-star properties, we separately investigate the impacts of crust and core segments. By matching the same core EOS to different crust EOSs, it is found that neutron-star radii are significantly affected by the crust segment. On the other hand, the neutron-star radii are also strongly dependent on the core EOS. However, the correlation between the radius and the symmetry energy slope of the core EOS is opposite to that of the crust EOS. It is likely that the nuclear model with a small slope parameter is favored by recent astrophysical observations.

nucl-th↗