ECCV 2026 论文列表

Cloud removal (CR) is essential for optical remote sensing,serving as a prerequisite for reliable downstream interpretation, suchas semantic segmentation and change detection. However, existing CRapproaches often prioritize visual realism while overlooking their im-pact on subsequent analytical tasks, leading to semantic drift and de-graded downstream performance. To address this issue, we propose Geo-Anchored Cloud Removal (GACR), a unified framework that jointlyensures faithful reconstruction and robust interpretability. At its core,GACR incorporates Observation-Anchored Residual Flow (OAR-Flow),which reformulates CR as a physically grounded residual inversion pro-cess. By anchoring the generative trajectory to the cloudy observationrather than pure noise, OAR-Flow enables fast, stable, and faithful re-construction. To further preserve semantic structures critical for down-stream interpretation, GACR integrates Geo-Contextual Prior Align-ment (GCPA) to constrain the reconstruction within a semantic manifoldinduced by a Vision Foundation Model (VFM). Consequently, GACRstrictly maintains the spatial-semantic integrity of complex landscapes.Extensive experiments across six CR datasets and twelve downstreamtasks demonstrate that GACR produces superior reconstruction qual-ity while consistently improving downstream task accuracy. The code isavailable at https://github.com/wzy6055/GACR.

Recent breakthroughs in 3D generation have advanced no-tably with the development of text-to-image diffusion model. However,existing methods remain two practical challenges: (1) They primarilygenerate single 3D object, but struggle to generate multi-object compo-sitional 3D assets due to the lack of the modeling for Gaussian prim-itives in reasonable interactions. (2) They often suffer from cross-viewinconsistency during 3D optimization, as Score Distillation Samplinginherently performs on each single view, inevitably resulting in cross-view hallucinations. To solve above issues, we propose I2 C-3D, a noveloptimization-based method to generate multi-view consistent composi-tional 3D assets with reasonable interactions. Specifically, we proposean Inclusive Interactive Collisions strategy to guide Gaussian primitivesappearing in reasonable interaction regions naturally, thereby ensuringobjects in the compositional scene interact in a physically plausible andvisually coherent way. Additionally, to enhance multi-view consistency,Multi-View Adaptive Score Distillation Sampling is devised to distillmulti-view consistency prior and layout prior from pre-trained diffusionmodel by modulating attention map of instance token and spatial to-ken across viewpoints. Benefiting from above elaborate designs, I2 C-3Dnot only generates high-fidelity multi-view consistent compositional 3Dassets but also supports 3D editing flexibly, facilitating complex scenegeneration. Extensive experiments demonstrate our I2 C-3D outperformsexisting methods in generation quality and multi-view consistency.

In this work, we propose Image-to-Image Rectified Flow Re-formulation (I2I-RFR), a practical plug-in reformulation that recastsstandard I2I regression networks as continuous-time transport models.While pixel-wise I2I regression is simple, stable, and easy to adapt acrosstasks, it often over-smooths ill-posed and multi-modal targets, whereasgenerative alternatives often require additional components, task-specifictuning, and more complex training and inference pipelines. Our methodaugments the backbone input by channel-wise concatenation with a noise-corrupted version of the ground-truth target and optimizes a simplet-reweighted pixel loss. This objective admits a rectified-flow interpre-tation via an induced velocity field, enabling ODE-based progressiverefinement at inference time while largely preserving the standard su-pervised training pipeline. In most cases, adopting I2I-RFR requiresonly expanding the input channels, and inference can be performed witha few explicit solver steps (e.g., 3 steps) without distillation. Exten-sive experiments across multiple image-to-image translation and videorestoration tasks show broad perceptual improvements, with possibleperception–distortion trade-offs in highly pixel-aligned settings such assuper-resolution.

We investigate the token mixer in vision backbones by revisit-ing clustering, one of the most classic approaches in machine learning. Aneffective token mixer is a fundamental component of modern vision back-bones like vision Transformers, facilitating information exchange betweenimage patches. Mainstream token mixers, which rely on convolution, at-tention, MLP, or their hybrids, primarily focus on navigating the trade-offbetween accuracy and computational cost. However, a significant draw-back of these methods is their black-box nature; their encoding processis opaque and lacks interpretability. Diverging from these opaque designs,we introduce ClusterMixer, a transparent token mixer that is grounded ina clustering paradigm and interpretable by design. ClusterMixer explicitlyformulates the token mixing process through a hierarchical clusteringmechanism. To model the natural, tree-like relationships inherent in visualdata, the clustering is performed in hyperbolic space, which is well-suitedfor embedding hierarchies with low distortion. Building on this innovation,we present HCFormer, a new backbone architecture that integratesClusterMixer with a series of meticulously designed clustering strategiesto ensure robust performance across tasks. Extensive experiments demon-strate that HCFormer consistently outperforms its counterparts acrossdiverse tasks, including image classification, object detection, instancesegmentation, and semantic segmentation. Considering its transparencyand efficacy, we hope HCFormer can facilitate a paradigm shift towardinterpretable backbones.

Monocular 3D lane detection plays a critical role in autonomous driv-ing, yet recovering reliable 3D geometry from a single image remains challeng-ing due to inherent depth ambiguity. Prior methods project image features intoBird’s-Eye-View (BEV) space under a flat-ground assumption, causing geomet-ric distortion on real-world roads. Recent methods instead predict explicit heightmaps to capture non-planar surfaces, but still rely on sparse anchor-based regres-sion and exploit the recovered geometry merely for spatial transformation ratherthan semantic understanding. To overcome these limitations, we propose HSDF-Lane, which implicitly models the road surface as a Height-aligned Signed Dis-tance Field (HSDF) over a densely sampled 3D feature volume. Through dif-ferentiable rendering, the HSDF jointly produces an accurate height map andsurface-aligned features. We further introduce Lane-aware Semantic PositionalEncoding (LSPE), which injects a lane-existence prior derived from the surface-aligned features into the transformer queries, coupling geometric structure withsemantic guidance. Extensive experiments on the OpenLane benchmark showthat HSDF-Lane achieves state-of-the-art performance in both 3D lane detectionand height map estimation. The code is available at https://github.com/JiyongBoo/HSDF-Lane.

Zero-shot object counting aims to count objects in images from givenclass names. Existing methods typically select representative image patches orextract text embeddings as exemplars to match image regions for counting. How-ever, a limited number of selected patches is insufficient to cover intra-class di-versity, and the accumulated errors in exemplar detection and selection substan-tially degrade counting accuracy. While the text embeddings remain at a genericsemantic level, blind to the specific instance diversity of target objects in animage. To handle these issues, this paper proposes the Hyper-Exemplar Repre-sentation (HER) counting framework (HER-Count) to represent the objects ofa specific category in a given image. Rather than image patches or text embed-dings, HERs are synthesized from holistic images and textual class names viaan multi-modal large-language model. By jointly integrating all target objects ineach image, HERs avoid exemplar generation errors and are more customizedrepresentative, thus yielding more accurate counting results. To effectively ex-ploit HERs in counting, HER-Count introduces a hierarchical fusion strategy thatinjects HERs into multiple intermediate layers of the vision encoder when pre-dicting density maps. This enables the model to progressively refine countingcues across layers, significantly enhancing the performance. A global discrimi-nation enhancement constraint at both image sample and semantic class levels isdesigned to further improve the capacities for both representation and discrimi-nation of HERs. Extensive experiments on widely used benchmarks demonstrateboth the state-of-the-art performance and generalization ability of the proposedHER-Count. Benefiting from the compact end-to-end design, HER-Count alsoachieves promising inference speed.

While Multimodal Large Language Models demonstrate im-pressive semantic capabilities, they often suffer from spatial blindness,struggling with fine-grained geometric reasoning and physical dynamics.Existing solutions typically rely on explicit 3D modalities or complexgeometric scaffolding, which are limited by data scarcity and generaliza-tion challenges. In this work, we propose a paradigm shift by leveragingthe implicit spatial prior within large-scale video generation models. Weposit that to synthesize temporally coherent videos, these models inher-ently learn robust 3D structural priors and physical laws. We introduceVEGA-3D (Video Extracted Generative Awareness), a plug-and-playframework that repurposes a pre-trained video diffusion model as a La-tent World Simulator. By extracting spatiotemporal features from inter-mediate noise levels and integrating them with semantic representationsvia a token-level adaptive gated fusion mechanism, we enrich MLLMswith dense geometric cues without explicit 3D supervision. Extensiveexperiments across 3D scene understanding, spatial reasoning, and em-bodied manipulation benchmarks demonstrate that our method outper-forms state-of-the-art baselines, validating that generative priors providea scalable foundation for physical-world understanding. Code is publiclyavailable at https://github.com/H-EmbodVis/VEGA-3D.

Recalling that diffusion models map noisy images to cleanones, the choice of noise addition (i.e., corruption) critically influencesthe denoising process. However, prior diffusion-based image restoration (IR)methods often employ naïve corruption strategies—such as randomlysampling noise, fixing the timestep, or even omitting corruption alto-gether—which may not accurately reflect the actual degradation of eachimage. To handle this, we propose Input-Aware Corruption for IR (IAC-IR) framework, which maps each low-quality (LQ) input into an optimalnoisy sample that lies on the pretrained diffusion trajectory. Instead ofchoosing corruption heuristically, we predict the timestep and noise us-ing supervision derived from the properties of the pretrained diffusionmodel. Specifically, the predicted timestep aligns the corrupted samplewith the Gaussian noise corrupted distribution, while the predicted noisepreserves the recoverable content of the input. Moreover, we use theseinput-aware corruption factors to improve conventional score-based dis-tillation. Rather than relying on random corruption, which often produceunreliable target scores and weak gradients, we perform distillation withinput-aware corruption, yielding more reliable score estimates and morestable distillation. By modeling input-aware corruption and integratingit into distillation, our method better leverages the pretrained diffusionprior, achieving superior perceptual quality in image restoration.

Real-world image restoration (RWIR) is a highly challeng-ing task due to the absence of clean ground-truth images. Many recentmethods resort to pseudo-label (PL) supervision, often within a Mean-Teacher (MT) framework, where a teacher network generates targets fora student network. However, these methods face a critical paradox: un-conditionally trusting the often imperfect, low-quality PLs forces the stu-dent model to learn undesirable artifacts, while discarding them severelylimits data diversity and impairs model generalization. In this paper, wepropose QualiTeacher, a novel framework that transforms pseudo-labelquality from a noisy liability into a conditional supervisory signal. In-stead of filtering, QualiTeacher explicitly conditions the student modelon the quality of the PLs, estimated by an ensemble of complementarynon-reference image quality assessment (NR-IQA) models spanning low-level distortion and semantic-level assessment. This strategy teaches thestudent network to learn a quality-graded restoration manifold, enablingit to understand what constitutes different quality levels. Consequently,it can not only avoid mimicking artifacts from low-quality labels but alsoextrapolate to generate results of higher quality than the teacher itself.To ensure the robustness and accuracy of this quality-driven learning,we further enhance the process with a multi-augmentation scheme to di-versify the PL quality spectrum, a score-based preference optimizationstrategy inspired by Direct Preference Optimization (DPO) to enforcea monotonically ordered quality separation, and a cropped consistencyloss to prevent adversarial over-optimization (reward hacking) of theIQA models. Experiments on standard RWIR benchmarks demonstratethat QualiTeacher can serve as a plug-and-play strategy to improve thequality of the existing pseudo-labeling framework, establishing a newparadigm for learning from imperfect supervision. Code will be releasedat https://github.com/fengyang1399-pixel/QualiTeacher.git .6.87.52 F. Xiao et al.(a) Unconditional Trust (b) Aggressive Filtering (c) QualiTeacher5.2 6.8 7.5 5.2 6.8 7.5Aug AugTeacher Teacher NR-IQA Teacher NR-IQAInput PL Input Optimal PL Input Quality-Conditioned PLsArtifacts are replicated rather than removed Limited data diversity, over-smoothed and blurry Better generalization via quality-conditioned pseudo-labelsDegraded Unconditional Trust Aggressive Filtering QualiTeacherFig. 1: Comparison of pseudo-label (PL) utilization strategies in the mean-teacherframework. (a) Unconditional Trust: PLs are used without filtering, causing degra-dation artifacts to be replicated by the student. (b) Aggressive Filtering: Low-quality PLs are discarded via NR-IQA filtering, yet over-smoothed outputs that receivedeceptively high scores survive, introducing blurriness. (c) QualiTeacher (Ours):Nearly all PLs are retained (with only extreme outliers discarded), with their qualityscores injected as continuous conditioning signals, enabling the student to leverage fulldata diversity while remaining aware of each sample’s reliability.

Visual intelligence requires anticipating the future behaviorof agents, yet vision systems lack a general representation for motion andbehavior. We propose dense point trajectories as visual tokens for behavior,a structured mid-level representation that disentangles motion fromappearance and generalizes across diverse non-rigid agents, such as animalsin-the-wild. Building on this abstraction, we design a diffusion transformerthat models unordered sets of trajectories and explicitly reasons aboutocclusion, enabling coherent forecasts of complex motion patterns. Toevaluate at scale, we curate 300 hours of unconstrained animal motion fromvideo through robust shot detection and camera-motion compensation.Experiments show that forecasting trajectory tokens achieves category-agnostic, data-efficient prediction, outperforms state-of-the-art baselines,and generalizes to rare species and morphologies, providing a foundationfor predictive visual intelligence in the wild.

Accurate reconstruction and tracking of dynamic human facesfrom image sequences is challenging because non-rigid deformations, ex-pression changes, and viewpoint variations occur simultaneously, creat-ing significant ambiguity in geometry and correspondence estimation. Wepresent a unified method for high-fidelity 4D facial reconstruction basedon canonical facial point prediction, a representation that assigns eachpixel a normalized facial coordinate in a shared canonical space. Thisformulation transforms dense tracking and dynamic reconstruction intoa canonical reconstruction problem, enabling temporally consistent ge-ometry and reliable correspondences within a single feed-forward model.By jointly predicting depth and canonical coordinates, our method en-ables accurate depth estimation, temporally stable reconstruction, dense3D geometry, and robust facial point tracking within a single architec-ture. We implement this formulation using a transformer-based modelthat jointly predicts depth and canonical facial coordinates, trained us-ing multi-view geometry data that non-rigidly warps into the canonicalspace. Extensive experiments on image and video benchmarks demon-strate state-of-the-art performance across reconstruction and trackingtasks, achieving approximately 3× lower correspondence error and fasterinference than prior dynamic reconstruction methods, while improvingdepth accuracy by 16%. These results highlight canonical facial point pre-diction as an effective foundation for unified feed-forward 4D facial recon-struction. Website: https://kocasariumut.github.io/FaceAnything/

Surface geometry and multispectral reflectance estimation are essential tools for reliable physical scene understanding in applications such as industrial inspection, cultural heritage analysis, and remote sensing. To obtain dependable spectral and geometric information, conventional pipelines often require multiple multispectral images captured across viewpoints, or multispectral photometric stereo with repeated, sequential illuminations. These approaches are frequently bandwidthintensive and slow to acquire due to their staged capture protocols. We propose EventSpecPS, an event-based multispectral photometric stereo system. It comprises a Spectrally-Angularly Multiplexed Encoding Device that encodes geometric and spectral cues into the event stream, which are then decoded by an Event-based Spectral-Angular Decoder to jointly recover surface normals and absolute multispectral reflectance. This enables single-stage, high-speed, and low-bandwidth acquisition and reconstruction, while remaining robust to cast shadows and specularities in practice. Extensive experiments on both synthetic and real data demonstrate state-of-the-art performance over existing methods. Code and data are available at https://github.com/Wujingqian/EventSpecPS.

While image stylization has been studied extensively, video stylization remains a critical and largely unsolved challenge in the field of intelligent content creation. Existing methods, usually utilizing a reference image as the style prior, suffer from content leakage, data scarcity and limited adaptability to long videos, leading to suboptimal results with severe style drift and motion distortion. For these issues, we present EchoStyle, a scalable text-driven framework to achieve high-quality stylization of videos with arbitrary lengths. To start with, we construct a video-to-video architecture to appropriately re-fuse the video content and the text style. To address data scarcity, we pioneer an automatic reverse-synthesis pipeline to establish V-Style20k, a large-scale stylization dataset of 20k high-quality video pairs. To facilitate long video stylization, we devise an init-follow-mode mechanism along with a slidingwindow inference strategy. Extensive experiments demonstrate EchoStyle’s excellent performance across a wide range of artistic styles, even comparable to leading closed-source solutions.

In this paper, we tackle the problem of performing consistent,unified modifications to a multi-shot video sequence. This task is par-ticularly challenging because multi-shot videos consist of discontinuoustemporal segments that vary significantly in viewpoint, camera scale, andsubject pose, leading to severe identity drift and cumulative error propa-gation. Achieving coherent edits requires establishing reliable cross-shotsemantic awareness to maintain stable subject appearance and visualcontinuity across these disjointed boundaries. To address this, we proposeMSEditor, the first framework designed specifically for consistent multi-shot video editing. To overcome the scarcity of high-quality multi-shottraining data, we repurpose existing multi-view video datasets to providerobust cross-shot supervision. Architecturally, we introduce a Supervi-sory Adapter that injects this cross-shot information into the diffusionbackbone, enabling the model to learn identity-consistent representa-tions. Furthermore, to effectively mitigate cumulative errors and ensurelong-range temporal coherence, we design a Cross-Shot Packing strat-egy that dynamically aggregates information from semantically relatedshots within the self-attention window. Extensive experiments demon-strate that MSEditor significantly outperforms existing methods on ourcurated multi-shot video editing benchmark in terms of identity preser-vation, temporal stability, and overall visual quality.

Previous works based on Segment Anything Model (SAM)have achieved promising performance in unified scene text detection andlayout analysis. However, the typical reliance on pixel-level text segmen-tation for sampling thousands of foreground points as prompts leadsto unsatisfied inference latency and limited data utilization. To addressabove issues, we propose ET-SAM, an Efficient framework with twodecoders for unified scene Text detection and layout analysis based onSAM. Technically, we customize a lightweight point decoder that pro-duces word heatmaps for achieving a few foreground points, thereby elim-inating excessive point prompts and accelerating inference. Without thedependence on pixel-level segmentation, we further design a joint train-ing strategy to leverage existing data with heterogeneous text-level an-notations. Specifically, the datasets with multi-level, word-level only, andline-level only annotations are combined in parallel as a unified trainingset. For these datasets, we introduce three corresponding sets of learn-able task prompts in both the point decoder and hierarchical mask de-coder to mitigate discrepancies across datasets. Extensive experimentsdemonstrate that, compared to the previous SAM-based architecture,ET-SAM achieves about 3× inference acceleration while obtaining com-petitive performance on HierText, and improves an average of 11.0% F-score on Total-Text, CTW1500, and ICDAR2015. The code and modelsare available at ET-SAM.

Transformer has recently garnered significant interest in un-supervised domain adaptation tasks due to its superior generalizationability. State-of-the-art methods leverage masked image modeling withconsistency regularization to improve target domain performance. How-ever, such strategy becomes less effective when the domain gap becomeslarge, as the domain gap between masked domain and original domainis not minimized. How to address the adaptation problem when domaingap becomes large is an important research problem in domain adap-tation. In this paper, we propose Dual Masked Generative AdversarialTransformer (Dual-MGAT), which simultaneously aligns the maskedtarget and masked source domain towards source and target domainto address the large domain gap problem. Specifically, we formulate ageneral framework for masked domain learning. To reduce the domaingap between the masked domain (e.g. masked target) and the originaldomain (e.g. source), we note that there exists two types of discrepan-cies, namely the information gap due to masked image modeling and thedomain distribution discrepancy. To this end, we propose the maskedgenerative adversarial adaptation technique, which introduces a trans-former decoder to bridge the masked feature space towards the originalfeature space via [CLS] token feature generation and utilizes a class con-ditional domain discriminator for adversarial distribution alignment toreduce the domain distribution discrepancy. We further investigate thedual branch of masked source to target domain adaptation, which boostsperformance. We provide a theoretical analysis of our framework from amasked domain adaptation perspective. Extensive experimental studiesdemonstrate the superiority of our framework.

Document layout understanding has long relied on bound-ing boxes with non-hierarchical and fixed label spaces. This abstractionfails on irregular regions, discards structural hierarchy, and falls shortof what real-world document tasks demand. We present DocLayout-VL, a foundation model that reframes document layout understandingas promptable open-set hierarchical segmentation. DocLayout-VL pro-duces a hierarchical layout tree, a pixel-accurate mask for each region,and a natural language explanation. Our novel Hierarchical GroundingContextualizer introduces Hierarchical Contextual Attention which mod-els inter-region relationships across hierarchical levels. DocLayout-VLis instruction-tuned on diverse layout tasks and domain-specific ontolo-gies. This enables highly customizable layout understanding via natu-ral language prompts, without any task-specific fine-tuning. To supportfoundation-scale training, we introduce HOPLay-Docs-80M, a large-scale dataset of 2.5M pages, 80M pixel-level masks, 400k open-set labelsand 5M promptable annotations across 12 domains and 58 documenttypes. We also release HOPLay-Docs-Bench, a human-verified evalu-ation suite covering open-set, hierarchical, and promptable layout seg-mentation. DocLayout-VL achieves state-of-the-art results across allbenchmarks, outperforming both specialized layout models and strongmultimodal baselines. Our work establishes a new paradigm for control-lable, semantically grounded layout understanding.

Open-vocabulary semantic segmentation (OVSS) leveragestextual semantics to segment objects beyond predefined categories. Whilethe self-supervised model DINOv3 provides strong structured visual rep-resentations, its lack of native textual alignment hinders direct applica-tion to OVSS. To bridge this gap, we propose DINOde, an ODE-basedframework that continuously aligns CLIP text embeddings to the DINOvisual manifold. Our approach employs two complementary components:(i) Semantic Text Flow (STF), which evolves text embeddings toward theDINO manifold through a continuous ODE trajectory, and (ii) GlobalContext Flow (GCF), which progressively refines the holistic image rep-resentation carried by DINO’s CLS token. To preserve the hypersphericalgeometry of the feature space during this evolution, we further intro-duce Velocity Tangent Projection, which constrains the learned veloc-ity field to the tangent space through projection. By modeling align-ment as a continuous trajectory, DINOde avoids the manifold entan-glement inherent in discrete MLP projections and yields more robustcross-modal alignment. Extensive experiments demonstrate that DIN-Ode consistently outperforms existing methods and achieves state-of-the-art performance across multiple OVSS benchmarks. The code is availableat https://github.com/yoon307/DINOde.

While large language models (LLMs) have advanced chestCT report generation, existing methods typically encode 3D volumesholistically, failing to distinguish informative cues from redundant anatom-ical background. Inspired by radiological cognitive subtraction, we pro-pose Differential Visual Prompting (DiffVP), which conditions reportgeneration on explicit, high-level semantic scan-to-reference differencesrather than solely on absolute visual features. DiffVP employs a hier-archical difference extractor to capture complementary global and localsemantic discrepancies into a shared latent space, along with a difference-to-prompt generator that transforms these signals into learnable visualprefix tokens for LLM conditioning. These difference prompts serve asstructured conditioning signals that implicitly suppress invariant anatomywhile amplifying diagnostically relevant visual evidence, thereby facil-itating accurate report generation without explicit lesion localization.On two large-scale benchmarks, DiffVP consistently outperforms priormethods, improving the average BLEU-1-4 by +10.98 and +4.36, respec-tively, and further boosts clinical efficacy on RadGenome-ChestCT (F1score 0.421). Code will be available at https://github.com/ArielTYH/DiffVP/.