arXiv subjects

Kangfu Mei

Publications and source records attributed to Kangfu Mei.

At least 19 recordsLinked to original sources

arXiv2026

XYZFlow:Scaling Multi dimensional Shortcut Flows for Efficient Generative Modeling

High-fidelity image generation faces a trade-off between speed and quality. Diffusion models produce strong visuals but require costly iterative sampling. Existing efficient methods mainly distill pretrained models into few-step samplers, a challenging process that depends heavily on teacher-model quality. In this paper, we introduce XYZFlow, a framework that rethinks efficient generation through multidimensional scaling of flow matching. Unlike single-step mappings, XYZFlow enhances expressivity by making probability paths more identifiable and learnable through structured multidimensional conditioning. We view autoregressive modeling as implicit flow straightening, where richer context reduces trajectory ambiguity. XYZFlow realizes this idea through two orthogonal dimensions: temporal scaling, which uses non-Markovian conditioning on the full denoising history; and spatial scaling, enabled by Next Shortcut Prediction, which sequentially generates patches using preceding patches' denoising trajectories as priors. Experiments show that XYZFlow achieves state-of-the-art performance, with 7.2-8.5X teacher speedups and competitive FID, while Next Shortcut Prediction delivers superior quality-latency trade-offs over model scaling or step reduction.

cs.CV↗

arXiv2026

SymbOmni: Evolving Agentic Omni Models via Symbolic Concept Learning

Visual generation is increasingly ubiquitous in diverse domains, from text-to-image/video synthesis to multimodal interactive creation. Yet prevailing monolithic models remain fundamentally constrained by their inability to learn cumulatively and evolve autonomously, which is a limitation we term the "perpetual novice" problem. They lack mechanisms for structuring experience into reusable knowledge and therefore rely on brittle, "from-scratch" reasoning for each task, resulting in poor compositional generalization and inefficient knowledge retention. Motivated by these limitations, we propose SymbOmni, an agentic omni-model designed for cumulative evolution through Symbolic Concept Learning. At its core is the Symbolic Concept Box, an optimizable memory module that abstracts low-level operations into reusable Symbolic Workflow Instructions. SymbOmni operates through an induction-transduction cycle: experiences are abstracted into symbolic concepts (induction), which are then adaptively composed to solve novel tasks (transduction). The training is done by verbalized backpropagation with language-based feedback to enable continuous self-improvement without gradient-based model fine-tuning. Comprehensive experiments validate that (I) SymbOmni significantly outperforms existing agent-based systems for iterative creation and also surpasses closed-source models (e.g., Nano Banana, GPT-Image-1) in both image quality and task success rates; (II) SymbOmni effectively reduces token consumption by over 40% while maintaining competitive generation quality; and (III) SymbOmni enables effective continual learning by achieving cumulative gains across multiple online-learning benchmarks and setting a new state of the art.

cs.CV↗

arXiv2026

Streaming Autoregressive Video Generation via Diagonal Distillation

Large pretrained diffusion models have significantly enhanced the quality of generated videos, and yet their use in real-time streaming remains limited. Autoregressive models offer a natural framework for sequential frame synthesis but require heavy computation to achieve high fidelity. Diffusion distillation can compress these models into efficient few-step variants, but existing video distillation approaches largely adapt image-specific methods that neglect temporal dependencies. These techniques often excel in image generation but underperform in video synthesis, exhibiting reduced motion coherence, error accumulation over long sequences, and a latency-quality trade-off. We identify two factors that result in these limitations: insufficient utilization of temporal context during step reduction and implicit prediction of subsequent noise levels in next-chunk prediction (i.e., exposure bias). To address these issues, we propose Diagonal Distillation, which operates orthogonally to existing approaches and better exploits temporal information across both video chunks and denoising steps. Central to our approach is an asymmetric generation strategy: more steps early, fewer steps later. This design allows later chunks to inherit rich appearance information from thoroughly processed early chunks, while using partially denoised chunks as conditional inputs for subsequent synthesis. By aligning the implicit prediction of subsequent noise levels during chunk generation with the actual inference conditions, our approach mitigates error propagation and reduces oversaturation in long-range sequences. We further incorporate implicit optical flow modeling to preserve motion quality under strict step constraints. Our method generates a 5-second video in 2.61 seconds (up to 31 FPS), achieving a 277.3x speedup over the undistilled model.

cs.CV↗

arXiv2025

Generative Manifold Distillation: Aligning Restoration Trajectories with Natural Image Prior

Pre-trained image restoration models often fail on out-of-distribution (OOD) real-world degradations. Adapting to these domains is challenging as real-world data lacks paired ground truth, and unsupervised methods often require unstable architectural changes. We propose Generative Manifold Distillation (GMD), which reframes domain adaptation as geometric manifold alignment. GMD operates in a strictly unpaired setting, requiring only low-quality (LQ) target observations. By leveraging the flow-matching dynamics of a frozen text-to-image foundation model, GMD projects off-manifold restorations onto the natural image manifold to generate high-quality pseudo-targets. To ensure stability, a quality-gated manifold filter rejects off-manifold samples, while source-anchored trajectory regularization prevents error accumulation. Ultimately, GMD distills a powerful generative prior into an efficient restoration network. Experiments demonstrate that GMD seamlessly adapts to new distributions using only LQ inputs, drastically improving perceptual quality with zero architectural modifications or added inference latency.

cs.CV↗

arXiv2025

Kernel Density Steering: Inference-Time Scaling via Mode Seeking for Image Restoration

Diffusion models show promise for image restoration, but existing methods often struggle with inconsistent fidelity and undesirable artifacts. To address this, we introduce Kernel Density Steering (KDS), a novel inference-time framework promoting robust, high-fidelity outputs through explicit local mode-seeking. KDS employs an $N$-particle ensemble of diffusion samples, computing patch-wise kernel density estimation gradients from their collective outputs. These gradients steer patches in each particle towards shared, higher-density regions identified within the ensemble. This collective local mode-seeking mechanism, acting as "collective wisdom", steers samples away from spurious modes prone to artifacts, arising from independent sampling or model imperfections, and towards more robust, high-fidelity structures. This allows us to obtain better quality samples at the expense of higher compute by simultaneously sampling multiple particles. As a plug-and-play framework, KDS requires no retraining or external verifiers, seamlessly integrating with various diffusion samplers. Extensive numerical validations demonstrate KDS substantially improves both quantitative and qualitative performance on challenging real-world super-resolution and image inpainting tasks.

cs.CV↗

arXiv2025

Reference-Guided Identity Preserving Face Restoration

Preserving face identity is a critical yet persistent challenge in diffusion-based image restoration. While reference faces offer a path forward, existing reference-based methods often fail to fully exploit their potential. This paper introduces a novel approach that maximizes reference face utility for improved face restoration and identity preservation. Our method makes three key contributions: 1) Composite Context, a comprehensive representation that fuses multi-level (high- and low-level) information from the reference face, offering richer guidance than prior singular representations. 2) Hard Example Identity Loss, a novel loss function that leverages the reference face to address the identity learning inefficiencies found in the existing identity loss. 3) A training-free method to adapt the model to multi-reference inputs during inference. The proposed method demonstrably restores high-quality faces and achieves state-of-the-art identity preserving restoration on benchmarks such as FFHQ-Ref and CelebA-Ref-Test, consistently outperforming previous work.

cs.CV↗

arXiv2025

The Power of Context: How Multimodality Improves Image Super-Resolution

Single-image super-resolution (SISR) remains challenging due to the inherent difficulty of recovering fine-grained details and preserving perceptual quality from low-resolution inputs. Existing methods often rely on limited image priors, leading to suboptimal results. We propose a novel approach that leverages the rich contextual information available in multiple modalities -- including depth, segmentation, edges, and text prompts -- to learn a powerful generative prior for SISR within a diffusion model framework. We introduce a flexible network architecture that effectively fuses multimodal information, accommodating an arbitrary number of input modalities without requiring significant modifications to the diffusion process. Crucially, we mitigate hallucinations, often introduced by text prompts, by using spatial information from other modalities to guide regional text-based conditioning. Each modality's guidance strength can also be controlled independently, allowing steering outputs toward different directions, such as increasing bokeh through depth or adjusting object prominence via segmentation. Extensive experiments demonstrate that our model surpasses state-of-the-art generative SISR methods, achieving superior visual quality and fidelity. See project page at https://mmsr.kfmei.com/.

cs.CV↗

arXiv2024

Bigger is not Always Better: Scaling Properties of Latent Diffusion Models

We study the scaling properties of latent diffusion models (LDMs) with an emphasis on their sampling efficiency. While improved network architecture and inference algorithms have shown to effectively boost sampling efficiency of diffusion models, the role of model size -- a critical determinant of sampling efficiency -- has not been thoroughly examined. Through empirical analysis of established text-to-image diffusion models, we conduct an in-depth investigation into how model size influences sampling efficiency across varying sampling steps. Our findings unveil a surprising trend: when operating under a given inference budget, smaller models frequently outperform their larger equivalents in generating high-quality results. Moreover, we extend our study to demonstrate the generalizability of the these findings by applying various diffusion samplers, exploring diverse downstream tasks, evaluating post-distilled models, as well as comparing performance relative to training compute. These findings open up new pathways for the development of LDM scaling strategies which can be employed to enhance generative capabilities within limited inference budgets.

cs.CV↗

arXiv2024

CoDi: Conditional Diffusion Distillation for Higher-Fidelity and Faster Image Generation

Large generative diffusion models have revolutionized text-to-image generation and offer immense potential for conditional generation tasks such as image enhancement, restoration, editing, and compositing. However, their widespread adoption is hindered by the high computational cost, which limits their real-time application. To address this challenge, we introduce a novel method dubbed CoDi, that adapts a pre-trained latent diffusion model to accept additional image conditioning inputs while significantly reducing the sampling steps required to achieve high-quality results. Our method can leverage architectures such as ControlNet to incorporate conditioning inputs without compromising the model's prior knowledge gained during large scale pre-training. Additionally, a conditional consistency loss enforces consistent predictions across diffusion steps, effectively compelling the model to generate high-quality images with conditions in a few steps. Our conditional-task learning and distillation approach outperforms previous distillation methods, achieving a new state-of-the-art in producing high-quality images with very few steps (e.g., 1-4) across multiple tasks, including super-resolution, text-guided image editing, and depth-to-image generation.

cs.CV↗

arXiv2023

Latent Feature-Guided Diffusion Models for Shadow Removal

Recovering textures under shadows has remained a challenging problem due to the difficulty of inferring shadow-free scenes from shadow images. In this paper, we propose the use of diffusion models as they offer a promising approach to gradually refine the details of shadow regions during the diffusion process. Our method improves this process by conditioning on a learned latent feature space that inherits the characteristics of shadow-free images, thus avoiding the limitation of conventional methods that condition on degraded images only. Additionally, we propose to alleviate potential local optima during training by fusing noise features with the diffusion network. We demonstrate the effectiveness of our approach which outperforms the previous best method by 13% in terms of RMSE on the AISTD dataset. Further, we explore instance-level shadow removal, where our model outperforms the previous best method by 82% in terms of RMSE on the DESOBA dataset.

cs.CV↗

arXiv2023

T1: Scaling Diffusion Probabilistic Fields to High-Resolution on Unified Visual Modalities

Diffusion Probabilistic Field (DPF) models the distribution of continuous functions defined over metric spaces. While DPF shows great potential for unifying data generation of various modalities including images, videos, and 3D geometry, it does not scale to a higher data resolution. This can be attributed to the ``scaling property'', where it is difficult for the model to capture local structures through uniform sampling. To this end, we propose a new model comprising of a view-wise sampling algorithm to focus on local structure learning, and incorporating additional guidance, e.g., text description, to complement the global geometry. The model can be scaled to generate high-resolution data while unifying multiple modalities. Experimental results on data generation in various modalities demonstrate the effectiveness of our model, as well as its potential as a foundation framework for scalable modality-unified visual content generation.

cs.CV↗

arXiv2022

Bi-Noising Diffusion: Towards Conditional Diffusion Models with Generative Restoration Priors

Conditional diffusion probabilistic models can model the distribution of natural images and can generate diverse and realistic samples based on given conditions. However, oftentimes their results can be unrealistic with observable color shifts and textures. We believe that this issue results from the divergence between the probabilistic distribution learned by the model and the distribution of natural images. The delicate conditions gradually enlarge the divergence during each sampling timestep. To address this issue, we introduce a new method that brings the predicted samples to the training data manifold using a pretrained unconditional diffusion model. The unconditional model acts as a regularizer and reduces the divergence introduced by the conditional model at each sampling step. We perform comprehensive experiments to demonstrate the effectiveness of our approach on super-resolution, colorization, turbulence removal, and image-deraining tasks. The improvements obtained by our method suggest that the priors can be incorporated as a general plugin for improving conditional diffusion models.

cs.CV↗

arXiv2022

VIDM: Video Implicit Diffusion Models

Diffusion models have emerged as a powerful generative method for synthesizing high-quality and diverse set of images. In this paper, we propose a video generation method based on diffusion models, where the effects of motion are modeled in an implicit condition manner, i.e. one can sample plausible video motions according to the latent feature of frames. We improve the quality of the generated videos by proposing multiple strategies such as sampling space truncation, robustness penalty, and positional group normalization. Various experiments are conducted on datasets consisting of videos with different resolutions and different number of frames. Results show that the proposed method outperforms the state-of-the-art generative adversarial network-based methods by a significant margin in terms of FVD scores as well as perceptible visual quality.

cs.CV↗

arXiv2022

AT-DDPM: Restoring Faces degraded by Atmospheric Turbulence using Denoising Diffusion Probabilistic Models

Although many long-range imaging systems are designed to support extended vision applications, a natural obstacle to their operation is degradation due to atmospheric turbulence. Atmospheric turbulence causes significant degradation to image quality by introducing blur and geometric distortion. In recent years, various deep learning-based single image atmospheric turbulence mitigation methods, including CNN-based and GAN inversion-based, have been proposed in the literature which attempt to remove the distortion in the image. However, some of these methods are difficult to train and often fail to reconstruct facial features and produce unrealistic results especially in the case of high turbulence. Denoising Diffusion Probabilistic Models (DDPMs) have recently gained some traction because of their stable training process and their ability to generate high quality images. In this paper, we propose the first DDPM-based solution for the problem of atmospheric turbulence mitigation. We also propose a fast sampling technique for reducing the inference times for conditional DDPMs. Extensive experiments are conducted on synthetic and real-world data to show the significance of our model. To facilitate further research, all codes and pretrained models are publically available at http://github.com/Nithin-GK/AT-DDPM

cs.CV↗

arXiv2022

Deep Semantic Statistics Matching (D2SM) Denoising Network

The ultimate aim of image restoration like denoising is to find an exact correlation between the noisy and clear image domains. But the optimization of end-to-end denoising learning like pixel-wise losses is performed in a sample-to-sample manner, which ignores the intrinsic correlation of images, especially semantics. In this paper, we introduce the Deep Semantic Statistics Matching (D2SM) Denoising Network. It exploits semantic features of pretrained classification networks, then it implicitly matches the probabilistic distribution of clear images at the semantic feature space. By learning to preserve the semantic distribution of denoised images, we empirically find our method significantly improves the denoising capabilities of networks, and the denoised results can be better understood by high-level vision tasks. Comprehensive experiments conducted on the noisy Cityscapes dataset demonstrate the superiority of our method on both the denoising performance and semantic segmentation accuracy. Moreover, the performance improvement observed on our extended tasks including super-resolution and dehazing experiments shows its potentiality as a new general plug-and-play component.

cs.CV↗

arXiv2022

A comparison of different atmospheric turbulence simulation methods for image restoration

Atmospheric turbulence deteriorates the quality of images captured by long-range imaging systems by introducing blur and geometric distortions to the captured scene. This leads to a drastic drop in performance when computer vision algorithms like object/face recognition and detection are performed on these images. In recent years, various deep learning-based atmospheric turbulence mitigation methods have been proposed in the literature. These methods are often trained using synthetically generated images and tested on real-world images. Hence, the performance of these restoration methods depends on the type of simulation used for training the network. In this paper, we systematically evaluate the effectiveness of various turbulence simulation methods on image restoration. In particular, we evaluate the performance of two state-or-the-art restoration networks using six simulations method on a real-world LRFID dataset consisting of face images degraded by turbulence. This paper will provide guidance to the researchers and practitioners working in this field to choose the suitable data generation models for training deep models for turbulence mitigation. The implementation codes for the simulation methods, source codes for the networks, and the pre-trained models will be publicly made available.

cs.CV↗

arXiv2022

Thermal to Visible Image Synthesis under Atmospheric Turbulence

In many practical applications of long-range imaging such as biometrics and surveillance, thermal imagining modalities are often used to capture images in low-light and nighttime conditions. However, such imaging systems often suffer from atmospheric turbulence, which introduces severe blur and deformation artifacts to the captured images. Such an issue is unavoidable in long-range imaging and significantly decreases the face verification accuracy. In this paper, we first investigate the problem with a turbulence simulation method on real-world thermal images. An end-to-end reconstruction method is then proposed which can directly transform thermal images into visible-spectrum images by utilizing natural image priors based on a pre-trained StyleGAN2 network. Compared with the existing two-steps methods of consecutive turbulence mitigation and thermal to visible image translation, our method is demonstrated to be effective in terms of both the visual quality of the reconstructed results and face verification accuracy. Moreover, to the best of our knowledge, this is the first work that studies the problem of thermal to visible image translation under atmospheric turbulence.

cs.CV↗

arXiv2021

LTT-GAN: Looking Through Turbulence by Inverting GANs

In many applications of long-range imaging, we are faced with a scenario where a person appearing in the captured imagery is often degraded by atmospheric turbulence. However, restoring such degraded images for face verification is difficult since the degradation causes images to be geometrically distorted and blurry. To mitigate the turbulence effect, in this paper, we propose the first turbulence mitigation method that makes use of visual priors encapsulated by a well-trained GAN. Based on the visual priors, we propose to learn to preserve the identity of restored images on a spatial periodic contextual distance. Such a distance can keep the realism of restored images from the GAN while considering the identity difference at the network learning. In addition, hierarchical pseudo connections are proposed for facilitating the identity-preserving learning by introducing more appearance variance without identity changing. Extensive experiments show that our method significantly outperforms prior art in both the visual quality and face verification accuracy of restored results.

cs.CV↗