arXiv subjects

Jinhua Zhang

Publications and source records attributed to Jinhua Zhang.

At least 19 recordsLinked to original sources

arXiv2026

Advanced Pixel Diffusion Model with Guided Sparse Global Refinement

Pixel-space diffusion has recently emerged as a promising direction for high-fidelity image generation by modeling images directly in the original pixel domain. However, pixel-space diffusion is computationally demanding due to the extremely high dimensionality of natural images. For efficiency, existing pixel diffusion models either compromise fine details with large-patch tokenization or confine subsequent refinement within individual patches. Such intra-patch refinement inevitably restricts structural continuity across patch boundaries and long-range token interactions, limiting refinement quality. To address these issues, we propose PixSGR, a novel Pixel diffusion framework with Sparse Global Refinement tailored for modeling the distribution of natural images directly in pixel space. PixSGR starts from a supervised low-channel bottleneck to efficiently capture the low-dimensional manifold of natural images. It then progressively expands the channel dimensionality and spatial resolution to recover increasingly fine-grained structures. At the spatial refinement stage, coarse-scale attention maps preselect globally relevant interactions to pre-sparsify fine-scale attention, enabling non-local refinement beyond isolated patches without the quadratic cost of dense attention. Extensive experiments on ImageNet validate the effectiveness of PixSGR. It achieves an FID of 1.51 at 256$\times$256 and maintains performance when scaled to 512$\times$512, attaining an FID of 1.60.

cs.CV↗

arXiv2026

Beyond Token-Level Cross-Entropy: Fréchet Distributional Post-Training for Autoregressive Image Generation

Autoregressive image generators are commonly pretrained with token-level cross-entropy under teacher forcing, yet evaluated by the distributional quality of decoded images. This creates an objective mismatch, because categorical errors have unequal image-level consequences, and a context mismatch, because inference conditions on model-generated histories. We introduce FD-loss post-training, which adapts a pretrained discrete generator using representation-space Fréchet distance as the sole objective. A dual-pass scheme first constructs detached rollout contexts through gradient-free generation under the model's native inference configuration, then performs differentiable replay with a probability-level straight-through estimator (STE) that preserves hard argmax decoding in the forward pass while propagating image-level gradients through temperature-scaled probabilities. Only the generator is updated, while the tokenizer and feature extractors remain frozen. Across eight completed configurations from four generator families on class-conditional ImageNet at $256\times256$, FD-loss post-training reduces FID and $\mathrm{FD}_{r6}$ by 41.4% and 52.0% on average. The strongest FID result improves from 2.42 to 1.43 without adding parameters or inference steps.

cs.CV↗

arXiv2026

IDESplat: Iterative Depth Probability Estimation for Generalizable 3D Gaussian Splatting

Generalizable 3D Gaussian Splatting aims to directly predict Gaussian parameters using a feed-forward network for scene reconstruction. Among these parameters, Gaussian means are particularly difficult to predict, so depth is usually estimated first and then unprojected to obtain the Gaussian sphere centers. Existing methods typically rely solely on a single warp to estimate depth probability, which hinders their ability to fully leverage cross-view geometric cues, resulting in unstable and coarse depth maps. To address this limitation, we propose IDESplat, which iteratively applies warp operations to boost depth probability estimation for accurate Gaussian mean prediction. First, to eliminate the inherent instability of a single warp, we introduce a Depth Probability Boosting Unit (DPBU) that integrates epipolar attention maps produced by cascading warp operations in a multiplicative manner. Next, we construct an iterative depth estimation process by stacking multiple DPBUs, progressively identifying potential depth candidates with high likelihood. As IDESplat iteratively boosts depth probability estimates and updates the depth candidates, the depth map is gradually refined, resulting in accurate Gaussian means. We conduct experiments on RealEstate10K, ACID, and DL3DV. IDESplat achieves outstanding reconstruction quality and state-of-the-art performance with real-time efficiency. On RE10K, it outperforms DepthSplat by 0.33 dB in PSNR, using only 10.7% of the parameters and 70% of the memory. Additionally, our IDESplat improves PSNR by 2.95 dB over DepthSplat on the DTU dataset in cross-dataset experiments, demonstrating its strong generalization ability.

cs.CV↗

arXiv2026

Texture Vector-Quantization and Reconstruction Aware Prediction for Generative Super-Resolution

Vector-quantized based models have recently demonstrated strong potential for visual prior modeling. However, existing VQ-based methods simply encode visual features with nearest codebook items and train index predictor with code-level supervision. Due to the richness of visual signal, VQ encoding often leads to large quantization error. Furthermore, training predictor with code-level supervision can not take the final reconstruction errors into consideration, result in sub-optimal prior modeling accuracy. In this paper we address the above two issues and propose a Texture Vector-Quantization and a Reconstruction Aware Prediction strategy. The texture vector-quantization strategy leverages the task character of super-resolution and only introduce codebook to model the prior of missing textures. While the reconstruction aware prediction strategy makes use of the straight-through estimator to directly train index predictor with image-level supervision. Our proposed generative SR model (TVQ&RAP) is able to deliver photo-realistic SR results with small computational cost.

cs.CV↗

arXiv2026

Taming Sampling Perturbations with Variance Expansion Loss for Latent Diffusion Models

Latent diffusion models have emerged as the dominant framework for high-fidelity and efficient image generation, owing to their ability to learn diffusion processes in compact latent spaces. However, while previous research has focused primarily on reconstruction accuracy and semantic alignment of the latent space, we observe that another critical factor, robustness to sampling perturbations, also plays a crucial role in determining generation quality. Through empirical and theoretical analyses, we show that the commonly used $β$-VAE-based tokenizers in latent diffusion models, tend to produce overly compact latent manifolds that are highly sensitive to stochastic perturbations during diffusion sampling, leading to visual degradation. To address this issue, we propose a simple yet effective solution that constructs a latent space robust to sampling perturbations while maintaining strong reconstruction fidelity. This is achieved by introducing a Variance Expansion loss that counteracts variance collapse and leverages the adversarial interplay between reconstruction and variance expansion to achieve an adaptive balance that preserves reconstruction accuracy while improving robustness to stochastic sampling. Extensive experiments demonstrate that our approach consistently enhances generation quality across different latent diffusion architectures, confirming that robustness in latent space is a key missing ingredient for stable and faithful diffusion sampling.

cs.CV↗

arXiv2026

MVAR: Visual Autoregressive Modeling with Scale and Spatial Markovian Conditioning

Essential to visual generation is efficient modeling of visual data priors. Conventional next-token prediction methods define the process as learning the conditional probability distribution of successive tokens. Recently, next-scale prediction methods redefine the process to learn the distribution over multi-scale representations, significantly reducing generation latency. However, these methods condition each scale on all previous scales and require each token to consider all preceding tokens, exhibiting scale and spatial redundancy. To better model the distribution by mitigating redundancy, we propose Markovian Visual AutoRegressive modeling (MVAR), a novel autoregressive framework that introduces scale and spatial Markov assumptions to reduce the complexity of conditional probability modeling. Specifically, we introduce a scale-Markov trajectory that only takes as input the features of adjacent preceding scale for next-scale prediction, enabling the adoption of a parallel training strategy that significantly reduces GPU memory consumption. Furthermore, we propose spatial-Markov attention, which restricts the attention of each token to a localized neighborhood of size k at corresponding positions on adjacent scales, rather than attending to every token across these scales, for the pursuit of reduced modeling complexity. Building on these improvements, we reduce the computational complexity of attention calculation from O(N^2) to O(Nk), enabling training with just eight NVIDIA RTX 4090 GPUs and eliminating the need for KV cache during inference. Extensive experiments on ImageNet demonstrate that MVAR achieves comparable or superior performance with both small model trained from scratch and large fine-tuned models, while reducing the average GPU memory footprint by 3.0x.

cs.CV↗

arXiv2026

Emotion and Acoustics Should Agree: Cross-Level Inconsistency Analysis for Audio Deepfake Detection

Audio Deepfake Detection (ADD) aims to detect spoof speech from bonafide speech. Most prior studies assume that stronger correlations within or across acoustic and emotional features imply authenticity, and thus focus on enhancing or measuring such correlations. However, existing methods often treat acoustic and emotional features in isolation or rely on correlation metrics, which overlook subtle desynchronization between them and smooth out abrupt discontinuities. To address these issues, we propose EAI-ADD, which treats cross level emotion acoustic inconsistency as the primary detection signal. We first project emotional and acoustic representations into a comparable space. Then we progressively integrate frame level and utterance level emotion features with acoustic features to capture cross level emotion acoustic inconsistencies across different temporal granularities. Experimental results on the ASVspoof 2019LA and 2021LA datasets demonstrate that the proposed EAI-ADD outperforms baselines, providing a more effective solution for audio anti spoofing detection.

cs.SD↗

arXiv2025

PerLDiff: Controllable Street View Synthesis Using Perspective-Layout Diffusion Models

Controllable generation is considered a potentially vital approach to address the challenge of annotating 3D data, and the precision of such controllable generation becomes particularly imperative in the context of data production for autonomous driving. Existing methods focus on the integration of diverse generative information into controlling inputs, utilizing frameworks such as GLIGEN or ControlNet, to produce commendable outcomes in controllable generation. However, such approaches intrinsically restrict generation performance to the learning capacities of predefined network architectures. In this paper, we explore the innovative integration of controlling information and introduce PerLDiff (\textbf{Per}spective-\textbf{L}ayout \textbf{Diff}usion Models), a novel method for effective street view image generation that fully leverages perspective 3D geometric information. Our PerLDiff employs 3D geometric priors to guide the generation of street view images with precise object-level control within the network learning process, resulting in a more robust and controllable output. Moreover, it demonstrates superior controllability compared to alternative layout control methods. Empirical results justify that our PerLDiff markedly enhances the precision of controllable generation on the NuScenes and KITTI datasets.

cs.CV↗

arXiv2025

Targeted False Positive Synthesis via Detector-guided Adversarial Diffusion Attacker for Robust Polyp Detection

Polyp detection is crucial for colorectal cancer screening, yet existing models are limited by the scale and diversity of available data. While generative models show promise for data augmentation, current methods mainly focus on enhancing polyp diversity, often overlooking the critical issue of false positives. In this paper, we address this gap by proposing an adversarial diffusion framework to synthesize high-value false positives. The extensive variability of negative backgrounds presents a significant challenge in false positive synthesis. To overcome this, we introduce two key innovations: First, we design a regional noise matching strategy to construct a negative synthesis space using polyp detection datasets. This strategy trains a negative-centric diffusion model by masking polyp regions, ensuring the model focuses exclusively on learning diverse background patterns. Second, we introduce the Detector-guided Adversarial Diffusion Attacker (DADA) module, which perturbs the negative synthesis process to disrupt a pre-trained detector's decision, guiding the negative-centric diffusion model to generate high-value, detector-confusing false positives instead of low-value, ordinary backgrounds. Our approach is the first to apply adversarial diffusion to lesion detection, establishing a new paradigm for targeted false positive synthesis and paving the way for more reliable clinical applications in colorectal cancer screening. Extensive results on public and in-house datasets verify the superiority of our method over the current state-of-the-arts, with our synthesized data improving the detectors by at least 2.6% and 2.7% in F1-score, respectively, over the baselines. Codes are at https://github.com/Huster-Hq/DADA.

cs.CV↗

arXiv2025

Generative Image Compression by Estimating Gradients of the Rate-variable Feature Distribution

While learned image compression (LIC) focuses on efficient data transmission, generative image compression (GIC) extends this framework by integrating generative modeling to produce photo-realistic reconstructed images. In this paper, we propose a novel diffusion-based generative modeling framework tailored for generative image compression. Unlike prior diffusion-based approaches that indirectly exploit diffusion modeling, we reinterpret the compression process itself as a forward diffusion path governed by stochastic differential equations (SDEs). A reverse neural network is trained to reconstruct images by reversing the compression process directly, without requiring Gaussian noise initialization. This approach achieves smooth rate adjustment and photo-realistic reconstructions with only a minimal number of sampling steps. Extensive experiments on benchmark datasets demonstrate that our method outperforms existing generative image compression approaches across a range of metrics, including perceptual distortion, statistical fidelity, and no-reference quality assessments.

eess.IV↗

arXiv2025

Full flexibility of entropies among ergodic measures for partially hyperbolic diffeomorphisms

We study nonhyperbolic and transitive partially hyperbolic diffeomorphisms having a one-dimensional center. We prove joint flexibility with respect to entropy and center Lyapunov exponent for a broad class of these systems. Flexibility means that for any given value of the center Lyapunov exponent and any value of entropy less than the supremum of entropies of ergodic measures with that exponent, there is an ergodic measure with exactly this entropy and exponent. Our hypotheses involve minimal foliations and blender-horseshoes, they formalize the interplay between two regions of the ambient space, one of center expanding and the other of center contracting type. The list of examples our results apply is rather long, a non-exhaustive list includes fibered by circles, flow-type, some Derived from Anosov diffeomorphisms, and some anomalous (non-dynamically coherent) diffeomorphisms.

math.DS↗

arXiv2025

A Conservative Partially Hyperbolic Dichotomy: Hyperbolicity versus Nonhyperbolic Measures

In a conservative and partially hyperbolic three-dimensional setting, we study three representative classes of diffeomorphisms: those homotopic to Anosov (or Derived from Anosov diffeomorphisms), diffeomorphisms in neighborhoods of the time-one map of the geodesic flow on a surface of negative curvature, and accessible and dynamically coherent skew products with circle fibers. In any of these classes, we establish the following dichotomy: either the diffeomorphism is Anosov, or it possesses nonhyperbolic ergodic measures. Our approach is perturbation-free and combines recent advances in the study of stably ergodic diffeomorphisms with a variation of the periodic approximation method to obtain ergodic measures. A key result in our construction, independent of conservative hypotheses, is the construction of nonhyperbolic ergodic measures for sets with a minimal strong unstable foliation that satisfy the mostly expanding property. This approach enables us to obtain nonhyperbolic ergodic measures in other contexts, including some subclasses of the so-called anomalous partially hyperbolic diffeomorphisms that are not dynamically coherent.

math.DS↗

arXiv2024

Transient infrared nanoscopy resolves the millisecond photoswitching dynamics of single lipid vesicles in water

Understanding the biophysical and biochemical properties of molecular nanocarriers under physiological conditions and with minimal interference is crucial for advancing nanomedicine, photopharmacology, drug delivery, nanotheranostics and synthetic biology. Yet, analytical methods struggle to combine precise chemical imaging and measurements without perturbative labeling. This challenge is exemplified for azobenzene-based photoswitchable lipids, which are intriguing reagents for controlling nanocarrier properties on fast timescales, enabling, e.g., precise light-induced drug release processes. Here, we leverage the chemical recognition and high spatio-temporal resolution of scattering-type scanning near-field optical microscopy (s-SNOM) to demonstrate non-destructive, label-free mid-infrared (MIR) imaging and spectroscopy of photoswitchable liposomes below the diffraction limit and the tracking of their dynamics down to 50 ms resolution. The vesicles are adsorbed on an ultrathin 10-nm SiN membrane, which separates the sample space from the tip space for stable and hour-long observations. By implementing a transient nanoscopy approach, we accurately resolve, for the first time, photoinduced changes in both the shape and the MIR spectral signature of individual vesicles and reveal abrupt change dynamics of the underlying photoisomerization process. Our findings highlight the methods potential for future studies on the complex dynamics of unlabeled nanoscale soft matter, as well as, in a broader context, for host-guest systems, energy materials or drugs.

physics.optics↗

arXiv2024

On physical measures of multi-singular hyperbolic vector fields

Bonatti and da Luz have introduced the class of \emph{multi-singular hyperbolic} vector fields to characterize systems whose periodic orbits and singularities do not bifurcate under perturbation (called star vector fields). In this paper, we study the Sina\"ı-Ruelle-Bowen measures for multi-singular hyperbolic vector fields: in a $C^1$ open and $C^1$ dense subset of multi-singular hyperbolic vector fields, each {$C^\infty$} one admits \emph{finitely} many physical measures whose basins cover a \emph{full} Lebesgue measure subset of the manifold. Similar results are also obtained for $C^1$ generic multi-singular hyperbolic vector fields.

math.DS↗

arXiv2024

The amount of nonhyperbolicity for partially hyperbolic diffeomorphisms

We study the amount of nonhyperbolicity within a broad class of (nonhyperbolic) partially hyperbolic diffeomorphisms with a one-dimensional center. For that, we focus on the center Lyapunov exponent and the entropy of its level sets. We show that these entropies vary continuously and can be expressed in terms of restricted variational principles. In this study, no dynamical coherence is required. Of particular interest is the case where the exponent is zero. To study this level set, we construct a compact set foliated by curves tangent to the central direction. Within this set, the entropy attains the maximal possible (and positive) value. Moreover, finite-time Lyapunov exponents converge uniformly to zero. In this construction, we introduce a mechanism to concatenate center curves. The class studied consists of those robustly transitive diffeomorphisms that have a pair of blender-horseshoes with different types of hyperbolicity and possess minimal strong stable and unstable foliations. This classes includes flow-type and circle-fibered diffeomorphisms as well as some derived from Anosov diffeomorphisms. It also includes the so-called anomalous examples which are dynamically incoherent.

math.DS↗

arXiv2024

Entropy properties of mostly expanding partially hyperbolic diffeomorphisms

The statistical properties of mostly expanding partially hyperbolic diffeomorphisms have been substantially studied. In this paper, we would like to address the entropy properties of mostly expanding partially hyperbolic diffeomorphisms. We prove that for mostly expanding partially hyperbolic diffeomorphisms with minimal strong stable foliation and one-dimensional center bundle, there exists a $C^1$-open neighborhood of them, in which the topological entropy varies continuously and the intermediate entropy property holds. To prove that, we show that each non-hyperbolic ergodic measure is approached by horseshoes in entropy and in weak$*$-topology.

math.DS↗

arXiv2023

Investigating 16 Open Clusters in the Kepler/K2-Gaia DR3 field. I. Membership, Binary, and Rotation

Using data from the Gaia Data Release 3 (Gaia DR3) and Kepler/K2, we present a catalog of 16 open clusters with ages ranging from 4 to 4000 Myr, which provides detailed information on membership, binary systems, and rotation. We assess the memberships in 5D phase space, and estimate the basic parameters of each cluster. Among the 20,160 members, there are 4,381 stars identified as binary candidates and 49 stars as blue straggler stars. The fraction of binaries vary in each cluster, and the range between 9% to 44%. We obtain the rotation periods of 5,467 members, of which 4,304 are determined in this work. To establish a benchmark for the rotation-age-color relation, we construct color-period diagrams. We find that the rotational features of binaries are similar to that of single stars, while features for binaries are more scattered in the rotation period. Moreover, the morphology of the color-period relationship is already established for Upper Scorpius at the age of 19 Myr, and some stars of varying spectral types (i.e. FG-, K-, and M-type) show different spin-down rates after the age of ~110 Myr. By incorporating the effects of stalled spin-down into our analysis, we develop an empirical rotation-age-color relation, which is valid with ages between 700 - 4000 Myr and colors corresponding to a range of 0.5 < (G_BP-G_RP)0 < 2.5 mag.

astro-ph.GA↗

arXiv2023

Betray Oneself: A Novel Audio DeepFake Detection Model via Mono-to-Stereo Conversion

Audio Deepfake Detection (ADD) aims to detect the fake audio generated by text-to-speech (TTS), voice conversion (VC) and replay, etc., which is an emerging topic. Traditionally we take the mono signal as input and focus on robust feature extraction and effective classifier design. However, the dual-channel stereo information in the audio signal also includes important cues for deepfake, which has not been studied in the prior work. In this paper, we propose a novel ADD model, termed as M2S-ADD, that attempts to discover audio authenticity cues during the mono-to-stereo conversion process. We first projects the mono to a stereo signal using a pretrained stereo synthesizer, then employs a dual-branch neural architecture to process the left and right channel signals, respectively. In this way, we effectively reveal the artifacts in the fake audio, thus improve the ADD performance. The experiments on the ASVspoof2019 database show that M2S-ADD outperforms all baselines that input mono. We release the source code at \url{https://github.com/AI-S2-Lab/M2S-ADD}.

cs.SD↗