arXiv subjects

Jun Yin

Publications and source records attributed to Jun Yin.

At least 73 records · Page 4Linked to original sources

arXiv2025

FloorplanMAE:A self-supervised framework for complete floorplan generation from partial inputs

In the architectural design process, floorplan design is often a dynamic and iterative process. Architects progressively draw various parts of the floorplan according to their ideas and requirements, continuously adjusting and refining throughout the design process. Therefore, the ability to predict a complete floorplan from a partial one holds significant value in the design process. Such prediction can help architects quickly generate preliminary designs, improve design efficiency, and reduce the workload associated with repeated modifications. To address this need, we propose FloorplanMAE, a self-supervised learning framework for restoring incomplete floor plans into complete ones. First, we developed a floor plan reconstruction dataset, FloorplanNet, specifically trained on architectural floor plans. Secondly, we propose a floor plan reconstruction method based on Masked Autoencoders (MAE), which reconstructs missing parts by masking sections of the floor plan and training a lightweight Vision Transformer (ViT). We evaluated the reconstruction accuracy of FloorplanMAE and compared it with state-of-the-art benchmarks. Additionally, we validated the model using real sketches from the early stages of architectural design. Experimental results show that the FloorplanMAE model can generate high-quality complete floor plans from incomplete partial plans. This framework provides a scalable solution for floor plan generation, with broad application prospects.

cs.AI↗

arXiv2025

Segment Any Architectural Facades (SAAF):An automatic segmentation model for building facades, walls and windows based on multimodal semantics guidance

In the context of the digital development of architecture, the automatic segmentation of walls and windows is a key step in improving the efficiency of building information models and computer-aided design. This study proposes an automatic segmentation model for building facade walls and windows based on multimodal semantic guidance, called Segment Any Architectural Facades (SAAF). First, SAAF has a multimodal semantic collaborative feature extraction mechanism. By combining natural language processing technology, it can fuse the semantic information in text descriptions with image features, enhancing the semantic understanding of building facade components. Second, we developed an end-to-end training framework that enables the model to autonomously learn the mapping relationship from text descriptions to image segmentation, reducing the influence of manual intervention on the segmentation results and improving the automation and robustness of the model. Finally, we conducted extensive experiments on multiple facade datasets. The segmentation results of SAAF outperformed existing methods in the mIoU metric, indicating that the SAAF model can maintain high-precision segmentation ability when faced with diverse datasets. Our model has made certain progress in improving the accuracy and generalization ability of the wall and window segmentation task. It is expected to provide a reference for the development of architectural computer vision technology and also explore new ideas and technical paths for the application of multimodal learning in the architectural field.

cs.CV↗

arXiv2025

FloorPlan-DeepSeek (FPDS): A multimodal approach to floorplan generation using vector-based next room prediction

In the architectural design process, floor plan generation is inherently progressive and iterative. However, existing generative models for floor plans are predominantly end-to-end generation that produce an entire pixel-based layout in a single pass. This paradigm is often incompatible with the incremental workflows observed in real-world architectural practice. To address this issue, we draw inspiration from the autoregressive 'next token prediction' mechanism commonly used in large language models, and propose a novel 'next room prediction' paradigm tailored to architectural floor plan modeling. Experimental evaluation indicates that FPDS demonstrates competitive performance in comparison to diffusion models and Tell2Design in the text-to-floorplan task, indicating its potential applicability in supporting future intelligent architectural design.

cs.CL↗

arXiv2025

MC$^2$A: Enabling Algorithm-Hardware Co-Design for Efficient Markov Chain Monte Carlo Acceleration

An increasing number of applications are exploiting sampling-based algorithms for planning, optimization, and inference. The Markov Chain Monte Carlo (MCMC) algorithms form the computational backbone of this emerging branch of machine learning. Unfortunately, the high computational cost limits their feasibility for large-scale problems and real-world applications, and the existing MCMC acceleration solutions are either limited in hardware flexibility or fail to maintain efficiency at the system level across a variety of end-to-end applications. This paper introduces \textbf{MC$^2$A}, an algorithm-hardware co-design framework, enabling efficient and flexible optimization for MCMC acceleration. Firstly, \textbf{MC$^2$A} analyzes the MCMC workload diversity through an extension of the processor performance roofline model with a 3rd dimension to derive the optimal balance between the compute, sampling and memory parameters. Secondly, \textbf{MC$^2$A} proposes a parametrized hardware accelerator architecture with flexible and efficient support of MCMC kernels with a pipeline of ISA-programmable tree-structured processing units, reconfigurable samplers and a crossbar interconnect to support irregular access. Thirdly, the core of \textbf{MC$^2$A} is powered by a novel Gumbel sampler that eliminates exponential and normalization operations. In the end-to-end case study, \textbf{MC$^2$A} achieves an overall {$307.6\times$, $1.4\times$, $2.0\times$, $84.2\times$} speedup compared to the CPU, GPU, TPU and state-of-the-art MCMC accelerator. Evaluated on various representative MCMC workloads, this work demonstrates and exploits the feasibility of general hardware acceleration to popularize MCMC-based solutions in diverse application domains.

cs.LG↗

arXiv2025

SAMST: A Transformer framework based on SAM pseudo label filtering for remote sensing semi-supervised semantic segmentation

Public remote sensing datasets often face limitations in universality due to resolution variability and inconsistent land cover category definitions. To harness the vast pool of unlabeled remote sensing data, we propose SAMST, a semi-supervised semantic segmentation method. SAMST leverages the strengths of the Segment Anything Model (SAM) in zero-shot generalization and boundary detection. SAMST iteratively refines pseudo-labels through two main components: supervised model self-training using both labeled and pseudo-labeled data, and a SAM-based Pseudo-label Refiner. The Pseudo-label Refiner comprises three modules: a Threshold Filter Module for preprocessing, a Prompt Generation Module for extracting connected regions and generating prompts for SAM, and a Label Refinement Module for final label stitching. By integrating the generalization power of large models with the training efficiency of small models, SAMST improves pseudo-label accuracy, thereby enhancing overall model performance. Experiments on the Potsdam dataset validate the effectiveness and feasibility of SAMST, demonstrating its potential to address the challenges posed by limited labeled data in remote sensing semantic segmentation.

cs.CV↗

arXiv2025

Localization-delocalization transition for a random block matrix model at the edge

Consider a random block matrix model consisting of $D$ random systems arranged along a circle, where each system is modeled by an independent $N\times N$ complex Hermitian Wigner matrix. Neighboring systems interact via an arbitrary deterministic $N\times N$ matrix $A$. In this paper, we extend the localization-delocalization transition previously established in arxiv:2312.07297 for the bulk eigenvalue spectrum to the entire spectrum, including the spectral edges. Let $[E^-,E^+]$ denote the support of the limiting spectral density, and define $κ_E:=|E-E^+|\wedge |E-E^-|$ as the distance from a given energy $E \in [E^-, E^+]$ to the spectral edges. We show that for eigenvalues near $E$, the corresponding eigenvectors undergo a localization-delocalization transition when $\|A\|_{\mathrm{HS}}$ crosses the critical threshold $(κ_E + N^{-2/3})^{-1/2}$. In the delocalized phase, the extreme eigenvalues asymptotically follow the Tracy-Widom distribution, while in the localized phase, the edge eigenvalue statistics asymptotically match those of $D$ independent GUE ensembles, up to a deterministic shift. Our results recover those of arxiv:2312.07297 in the bulk regime, where $κ_E \asymp 1$, and further reveal the presence of mobility edges near $E^\pm$ when $1 \ll \|A\|_{\mathrm{HS}} \ll N^{1/3}$. Specifically, bulk eigenvectors corresponding to energies $E$ with $κ_E \gg \|A\|_{\mathrm{HS}}^{-2}$ are delocalized, while those with $κ_E \ll \|A\|_{\mathrm{HS}}^{-2}$ are localized.

math.PR↗

arXiv2025

Delocalization of random band matrices at the edge

We consider $N\times N$ Hermitian random band matrices $H=(H_{xy})$, whose entries are centered complex Gaussian random variables. The indices $x,y$ range over the $d$-dimensional discrete torus $(\mathbb Z/L\mathbb Z)^d$ with $d\in \{1,2\}$ and $N=L^d$. The variance profile $S_{xy}=\mathbb E|h_{xy}|^2$ exhibits a banded structure: specifically, $S_{xy}=0$ whenever the distance $|x-y|$ exceeds a band width parameter $W\le L$. Let $W=L^α$ for some exponent $0<α\le 1$. We show that as $α$ increases from $\mathbf 1_{d=1}/2$ to $1-d/6$, the range of energies corresponding to delocalized eigenvectors gradually expands from the bulk toward the entire spectrum. More precisely, we prove that eigenvectors associated with energies $E$ satisfying $2 - |E| \gg N^{-c_{d,α}}$ are delocalized, where the exponent $c_{d,α}$ is given by $c_{d,α} = 2α- 1$ in dimension 1 and $c_{d,α} = α$ in dimension 2. Furthermore, when $α> 1-d/6$, all eigenvectors of $H$ become delocalized. We further establish quantum unique ergodicity for delocalized eigenvectors, as well as a rigidity estimate for the eigenvalues. Our findings extend previous results -- established in the bulk regime for one-dimensional (1D) (arXiv:2501.01718) and two-dimensional (2D) (arXiv:2503.07606) random band matrices -- to the entire spectrum, including the spectral edges. They also complement the results of arXiv:0906.4047 and arXiv:2401.00492, which concern the edge eigenvalue statistics for 1D and 2D random band matrices.

math.PR↗

arXiv2025

ArchShapeNet:An Interpretable 3D-CNN Framework for Evaluating Architectural Shapes

In contemporary architectural design, the growing complexity and diversity of design demands have made generative plugin tools essential for quickly producing initial concepts and exploring novel 3D forms. However, objectively analyzing the differences between human-designed and machine-generated 3D forms remains a challenge, limiting our understanding of their respective strengths and hindering the advancement of generative tools. To address this, we built ArchForms-4000, a dataset containing 2,000 architect-designed and 2,000 Evomass-generated 3D forms; Proposed ArchShapeNet, a 3D convolutional neural network tailored for classifying and analyzing architectural forms, incorporating a saliency module to highlight key spatial features aligned with architectural reasoning; And conducted comparative experiments showing our model outperforms human experts in distinguishing form origins, achieving 94.29% accuracy, 96.2% precision, and 98.51% recall. This study not only highlights the distinctive advantages of human-designed forms in spatial organization, proportional harmony, and detail refinement but also provides valuable insights for enhancing generative design tools in the future.

cs.CV↗

arXiv2025

On the localization length of finite-volume random block Schrödinger operators

We study a general class of random block Schrödinger operators (RBSOs) in dimensions 1 and 2, which naturally extend the Anderson model by replacing the random potential with a random block potential. Specifically, we focus on two RBSOs -- the block Anderson and Wegner orbital models -- defined on the $d$-dimensional torus $(\mathbb Z/L\mathbb Z)^d$. They take the form $H=V + λΨ$, where $V$ is a block potential with i.i.d. $W^d\times W^d$ Gaussian diagonal blocks, $Ψ$ describes interactions between neighboring blocks, and $λ>0$ is a coupling parameter. We normalize the blocks of $Ψ$ so that each block has a Hilbert-Schmidt norm of the same order as the blocks of $V$. Assuming $W\ge L^δ$ for a small constant $δ>0$ and $λ\gg W^{-d/2}$, we establish the following results. In dimension $d=2$, we prove delocalization and quantum unique ergodicity for bulk eigenvectors. Combined with the localization result from arXiv:1608.02922, which holds under the condition $λ\ll W^{-d/2}$, this provides a rigorous proof of the Anderson localization-delocalization transition as $λ$ crosses the critical threshold $W^{-d/2}$. In dimension $d=1$, we show that the localization length of bulk eigenvectors is at least of order $(Wλ)^2$, which is believed to be the correct scaling.

math.PR↗

arXiv2025

Delocalization of One-Dimensional Random Band Matrices

Consider an $ N \times N$ Hermitian one-dimensional random band matrix with band width $W > N^{1 / 2 + \frak c} $ for any $ {\frak c} > 0$. In the bulk of the spectrum and in the large $ N $ limit, we obtain the following results: (i) The semicircle law holds up to the scale $ N^{-1 + \varepsilon} $ for any $ \varepsilon > 0 $. (ii) All $ L^2 $- normalized eigenvectors are delocalized, meaning their $ L^\infty$ norms are simultaneously bounded by $ N^{-\frac{1}{2} + \varepsilon} $ with overwhelming probability, for any $ \varepsilon > 0 $. (iii) Quantum unique ergodicity holds in the sense that the local $ L^2 $ mass of eigenvectors becomes equidistributed with high probability. (iv) Universality of eigenvalue statistics holds, i.e., the local eigenvalue statistics of these band matrices are given by those of Gaussian unitary ensembles.

math.PR↗

arXiv2025

Delocalization of a general class of random block Schrödinger operators

We consider a natural class of extensions of the Anderson model on $\mathbb Z^d$, called random block Schrödinger operators (RBSOs), defined on the $d$-dimensional torus $(\mathbb Z/L\mathbb Z)^d$. These operators take the form $H=V+λΨ$, where $V$ is a diagonal block matrix whose diagonal blocks are i.i.d. $W^d\times W^d$ GUE, representing a random block potential, $Ψ$ describes interactions between neighboring blocks, and $λ\ll 1$ is a small coupling parameter (making $H$ a perturbation of $V$). We focus on three specific RBSOs: (1) the block Anderson model, where $Ψ$ is the discrete Laplacian on $(\mathbb Z/L\mathbb Z)^d$; (2) the Anderson orbital model, where $Ψ$ is a block Laplacian operator; (3) the Wegner orbital model, where the nearest-neighbor blocks of $Ψ$ are themselves random matrices. Assuming $d\ge 7$ and $W\ge L^\varepsilon$ for a small constant $\varepsilon>0$, and under a certain lower bound on $λ$, we establish delocalization and quantum unique ergodicity for bulk eigenvectors, along with quantum diffusion estimates for the Green's function. Combined with the localization results of arXiv:1608.02922, our results rigorously demonstrate the existence of an Anderson localization-delocalization transition for RBSOs as $λ$ varies. Our proof is based on the $T$-expansion method and the concept of self-energy renormalization, originally developed in the study of random band matrices in arXiv:2104.12048. In addition, we introduce a conceptually novel idea, called coupling renormalization, which extends the notion of self-energy renormalization. While this phenomenon is well-known in quantum field theory, it is identified here for the first time in the context of random Schrödinger operators. We expect that our methods can be extended to models with real or non-Gaussian block potentials, as well as more general forms of interactions.

math.PR↗

arXiv2025

Stacking-induced ferroelectricity in tetralayer graphene

Recent studies have reported emergent ferroelectric behavior in twisted or moiré-engineered graphene-based van der Waals heterostructures, yet the microscopic origin of this effect remains under debate. Pristine mono- or few-layer graphene lacks a permanent dipole due to its centrosymmetric lattice, making the emergence of ferroelectricity unlikely. However, mixed-stacked graphene, such as the ABCB tetralayer configuration, breaks both inversion and mirror symmetry and has been theoretically predicted to support electrically switchable dipoles. ABCB graphene represents the simplest natural graphene polytype exhibiting intrinsic out-of-plane polarization, arising from asymmetric charge carrier distribution across its layers. Here, we report robust ferroelectric behavior in dual-gated, non-aligned ABCB tetralayer graphene encapsulated in hexagonal boron nitride. The device exhibits pronounced hysteresis in resistance under both top and bottom gate modulation, with the effect persisting up to room temperature. This hysteresis originates from reversible layer-polarized charge reordering, driven by gate-induced transitions between ABCB and BCBA stacking configurations -- without requiring moiré superlattices. Our findings establish stacking-order-induced symmetry breaking as a fundamental route to electronic ferroelectricity in graphene and open pathways for non-volatile memory applications based on naturally occurring mixed-stacked multilayer graphene.

cond-mat.mes-hall↗

arXiv2025

TSCnet: A Text-driven Semantic-level Controllable Framework for Customized Low-Light Image Enhancement

Deep learning-based image enhancement methods show significant advantages in reducing noise and improving visibility in low-light conditions. These methods are typically based on one-to-one mapping, where the model learns a direct transformation from low light to specific enhanced images. Therefore, these methods are inflexible as they do not allow highly personalized mapping, even though an individual's lighting preferences are inherently personalized. To overcome these limitations, we propose a new light enhancement task and a new framework that provides customized lighting control through prompt-driven, semantic-level, and quantitative brightness adjustments. The framework begins by leveraging a Large Language Model (LLM) to understand natural language prompts, enabling it to identify target objects for brightness adjustments. To localize these target objects, the Retinex-based Reasoning Segment (RRS) module generates precise target localization masks using reflection images. Subsequently, the Text-based Brightness Controllable (TBC) module adjusts brightness levels based on the generated illumination map. Finally, an Adaptive Contextual Compensation (ACC) module integrates multi-modal inputs and controls a conditional diffusion model to adjust the lighting, ensuring seamless and precise enhancements accurately. Experimental results on benchmark datasets demonstrate our framework's superior performance at increasing visibility, maintaining natural color balance, and amplifying fine details without creating artifacts. Furthermore, its robust generalization capabilities enable complex semantic-level lighting adjustments in diverse open-world environments through natural language interactions.

cs.CV↗

arXiv2025

PromptLNet: Region-Adaptive Aesthetic Enhancement via Prompt Guidance in Low-Light Enhancement Net

Learning and improving large language models through human preference feedback has become a mainstream approach, but it has rarely been applied to the field of low-light image enhancement. Existing low-light enhancement evaluations typically rely on objective metrics (such as FID, PSNR, etc.), which often result in models that perform well objectively but lack aesthetic quality. Moreover, most low-light enhancement models are primarily designed for global brightening, lacking detailed refinement. Therefore, the generated images often require additional local adjustments, leading to research gaps in practical applications. To bridge this gap, we propose the following innovations: 1) We collect human aesthetic evaluation text pairs and aesthetic scores from multiple low-light image datasets (e.g., LOL, LOL2, LOM, DCIM, MEF, etc.) to train a low-light image aesthetic evaluation model, supplemented by an optimization algorithm designed to fine-tune the diffusion model. 2) We propose a prompt-driven brightness adjustment module capable of performing fine-grained brightness and aesthetic adjustments for specific instances or regions. 3) We evaluate our method alongside existing state-of-the-art algorithms on mainstream benchmarks. Experimental results show that our method not only outperforms traditional methods in terms of visual quality but also provides greater flexibility and controllability, paving the way for improved aesthetic quality.

cs.CV↗

arXiv2025

Delocalization of Two-Dimensional Random Band Matrices

We study a random band matrix $H=(H_{xy})_{x,y}$ of dimension $N\times N$ with mean-zero complex Gaussian entries, where $x,y$ belong to the discrete torus $(\mathbb{Z}/\sqrt{N}\mathbb{Z})^{2}$. The variance profile $\mathbb{E}|H_{xy}|^{2}=S_{xy}$ vanishes when the distance between $x,y$ is larger than some band-width parameter $W$ depending on $N$. We show that if the band-width satisfies $W\geq N^{\mathfrak{c}}$ for some $\mathfrak{c}>0$, then in the large-$N$ limit, we have the following results. The first result is a local semicircle law in the bulk down to scales $N^{-1+\varepsilon}$. The second is delocalization of bulk eigenvectors. The third is a quantum unique ergodicity for bulk eigenvectors. The fourth is universality of local bulk eigenvalue statistics. The fifth is a quantum diffusion profile for the associated $T$ matrix. Our method is based on embedding $H$ inside a matrix Brownian motion $H_{t}$ as done in [Dubova-Yang '24] and [Yau-Yin '25] for band matrices on the one-dimensional torus. In this paper, the key additional ingredient in our analysis of $H_{t}$ is a new CLT-type estimate for polynomials in the entries of the resolvent of $H_{t}$.

math.PR↗

arXiv2025

CNN-based Robust Sound Source Localization with SRP-PHAT for the Extreme Edge

Robust sound source localization for environments with noise and reverberation are increasingly exploiting deep neural networks fed with various acoustic features. Yet, state-of-the-art research mainly focuses on optimizing algorithmic accuracy, resulting in huge models preventing edge-device deployment. The edge, however, urges for real-time low-footprint acoustic reasoning for applications such as hearing aids and robot interactions. Hence, we set off from a robust CNN-based model using SRP-PHAT features, Cross3D [16], to pursue an efficient yet compact model architecture for the extreme edge. For both the SRP feature representation and neural network, we propose respectively our scalable LC-SRP-Edge and Cross3D-Edge algorithms which are optimized towards lower hardware overhead. LC-SRP-Edge halves the complexity and on-chip memory overhead for the sinc interpolation compared to the original LC-SRP [19]. Over multiple SRP resolution cases, Cross3D-Edge saves 10.32~73.71% computational complexity and 59.77~94.66% neural network weights against the Cross3D baseline. In terms of the accuracy-efficiency tradeoff, the most balanced version (EM) requires only 127.1 MFLOPS computation, 3.71 MByte/s bandwidth, and 0.821 MByte on-chip memory in total, while still retaining competitiveness in state-of-the-art accuracy comparisons. It achieves 8.59 ms/frame end-to-end latency on a Rasberry Pi 4B, which is 7.26x faster than the corresponding baseline.

eess.AS↗

arXiv2025

When Graph meets Multimodal: Benchmarking and Meditating on Multimodal Attributed Graphs Learning

Multimodal Attributed Graphs (MAGs) are ubiquitous in real-world applications, encompassing extensive knowledge through multimodal attributes attached to nodes (e.g., texts and images) and topological structure representing node interactions. Despite its potential to advance diverse research fields like social networks and e-commerce, MAG representation learning (MAGRL) remains underexplored due to the lack of standardized datasets and evaluation frameworks. In this paper, we first propose MAGB, a comprehensive MAG benchmark dataset, featuring curated graphs from various domains with both textual and visual attributes. Based on MAGB dataset, we further systematically evaluate two mainstream MAGRL paradigms: $\textit{GNN-as-Predictor}$, which integrates multimodal attributes via Graph Neural Networks (GNNs), and $\textit{VLM-as-Predictor}$, which harnesses Vision Language Models (VLMs) for zero-shot reasoning. Extensive experiments on MAGB reveal following critical insights: $\textit{(i)}$ Modality significances fluctuate drastically with specific domain characteristics. $\textit{(ii)}$ Multimodal embeddings can elevate the performance ceiling of GNNs. However, intrinsic biases among modalities may impede effective training, particularly in low-data scenarios. $\textit{(iii)}$ VLMs are highly effective at generating multimodal embeddings that alleviate the imbalance between textual and visual attributes. These discoveries, which illuminate the synergy between multimodal attributes and graph topologies, contribute to reliable benchmarks, paving the way for future MAG research. The MAGB dataset and evaluation pipeline are publicly available at https://github.com/sktsherlock/MAGB.

cs.LG↗

arXiv2025

A random matrix model towards the quantum chaos transition conjecture

Consider $D$ random systems that are modeled by independent $N\times N$ complex Hermitian Wigner matrices. Suppose they are lying on a circle and the neighboring systems interact with each other through a deterministic matrix $A$. We prove that in the asymptotic limit $N\to \infty$, the whole system exhibits a quantum chaos transition when the interaction strength $\|A\|_{HS}$ varies. Specifically, when $\|A\|_{HS}\ge N^{\varepsilon}$, we prove that the bulk eigenvalue statistics match those of a $DN\times DN$ GUE asymptotically and each bulk eigenvector is approximately equally distributed among the $D$ subsystems with probability $1-o(1)$. These phenomena indicate quantum chaos of the whole system. In contrast, when $\|A\|_{HS}\le N^{-\varepsilon}$, we show that the system is integrable: the bulk eigenvalue statistics behave like $D$ independent copies of GUE statistics asymptotically and each bulk eigenvector is localized on only one subsystem. In particular, if we take $D\to \infty$ after the $N\to \infty$ limit, the bulk statistics converge to a Poisson point process under the $DN$ scaling.

math.PR↗