| Total: 1212
Computing geodesic distances on 3D surfaces is fundamental to many tasks in 3D vision and geometry processing, with deep connections to tasks such as shape correspondence. Recent learning-based methods achieve strong performance but rely on large 3D backbones, leading to high memory usage and latency, which limit their use in interactive or resource-constrained settings. We introduce LiteGE, a lightweight approach that constructs compact, category-aware shape descriptors by applying PCA to unsigned distance field (UDFs) samples at informative voxels. This descriptor is efficient to compute and removes the need for high-capacity networks. LiteGE remains robust on sparse point clouds, supporting inputs with as few as 300 points, where prior methods fail. Extensive experiments show that LiteGE reduces memory usage and inference time by up to 300x compared to existing neural approaches. In addition, by exploiting the intrinsic relationship between geodesic distance and shape correspondence, LiteGE enables fast and accurate shape matching. Our method achieves up to 1000x speedup over state-of-the-art mesh-based approaches while maintaining comparable accuracy on non-isometric shape pairs, including evaluations on point-cloud inputs.
We introduce a new task of open-world object counting in videos: given a text description, or an image example, that specifies the target object, the objective is to enumerate all the unique instances of the target objects in the video. This task is especially challenging in crowded scenes with occlusions and objects of similar appearance, where avoiding double counting and identifying reappearances is crucial. To this end, we make the following contributions: we introduce a model, CountVid, for this task. It leverages an image-based counting model, and a promptable video segmentation and tracking model, to enable automated open-world object counting across video frames. To evaluate its performance, we introduce VideoCount, a new dataset for this novel task built from the TAO and MOT20 tracking datasets, as well as from videos of penguins and metal alloy crystallization captured by x-rays. Using this dataset, we demonstrate that CountVid provides accurate object counts, and significantly outperforms strong baselines.
Omnidirectional videos (ODVs) provide an immersive visual experience by capturing the 360° scene. With the rapid advancements in virtual/augmented reality, metaverse, and generative artificial intelligence, the demand for high-quality ODVs is surging. However, ODVs often suffer from low resolution due to their wide field of view and limitations in capturing devices and transmission bandwidth. Although video super-resolution (SR) is a capable video quality enhancement technique, the performance ceiling and practical generalization of existing methods are limited when applied to ODVs due to their unique attributes. To alleviate spatial projection distortions and temporal flickering of ODVs, we propose a Spatio-Temporal Distortion Aware Network (STDAN) with joint spatio-temporal alignment and reconstruction. Specifically, we incorporate a spatio-temporal continuous alignment (STCA) to mitigate discrete geometric artifacts in parallel with temporal alignment. Subsequently, we introduce an interlaced multi-frame reconstruction (IMFR) to enhance temporal consistency. Furthermore, we employ latitude-saliency adaptive (LSA) weights to focus on regions with higher texture complexity and human-watching interest. By exploring a spatio-temporal jointly framework and real-world viewing strategies, STDAN effectively reinforces spatio-temporal coherence on a novel ODV-SR dataset and ensures affordable computational costs. Extensive experimental results demonstrate that STDAN outperforms state-of-the-art methods in improving visual fidelity and dynamic smoothness of ODVs.
Multimodal Retrieval-Augmented Generation (MRAG) has recently been explored to empower Large Vision Language Models (LVLMs) with more comprehensive and up-to-date contextual knowledge, aiming to compensate for their limited and coarse-grained parametric knowledge in knowledge-intensive tasks. However, the retrieved contextual knowledge is usually not aligned with LVLMs’ internal parametric knowledge, leading to knowledge conflicts and further unreliable responses. To tackle this issue, we design KCM, a training-free and plug-and-play framework that can effectively mitigate knowledge conflicts while incorporating MRAG for more accurate LVLM responses. KCM enhances contextual knowledge utilization by modifying the LVLM architecture from three key perspectives. First, KCM adaptively adjusts attention distributions among multiple attention heads, encouraging LVLMs to focus on contextual knowledge with reduced distraction. Second, KCM identifies and prunes knowledge-centric LVLM neurons that encode coarse-grained parametric knowledge, thereby suppressing interferences and enabling more effective integration of contextual knowledge. Third, KCM amplifies the information flow from the input context by injecting supplementary context logits, reinforcing its contribution to the final output. Extensive experiments over multiple LVLMs and benchmarks show that KCM outperforms the state-of-the-art consistently by large margins, incurring neither extra training nor external tools.
Maintaining consistent 3D scene representations over time is a significant challenge in computer vision. Updating 3D scenes from sparse-view observations is crucial for various real-world applications, including urban planning, disaster assessment, and historical site preservation, where dense scans are often unavailable or impractical. In this paper, we propose Cross-Temporal 3D Gaussian Splatting (Cross-Temporal 3DGS), a novel framework for efficiently reconstructing and updating 3D scenes across different time periods, using sparse images and previously captured scene priors. Our approach comprises three stages: 1) Cross-temporal camera alignment for estimating and aligning camera poses across different timestamps; 2) Interference-based confidence initialization to identify unchanged regions between timestamps, thereby guiding updates; and 3) Progressive cross-temporal optimization, which iteratively integrates historical prior information into the 3D scene to enhance reconstruction quality. Our method supports non-continuous capture, enabling not only updates using new sparse views to refine existing scenes, but also recovering past scenes from limited data with the help of current captures. Furthermore, we demonstrate the potential of this approach to achieve temporal changes using only sparse images, which can later be reconstructed into detailed 3D representations as needed. Experimental results show significant improvements over baseline methods in reconstruction quality and data efficiency, making this approach a promising solution for scene versioning, cross-temporal digital twins, and long-term spatial documentation.
Recent camera-based 3D semantic scene completion (SSC) methods have increasingly explored leveraging temporal cues to enrich the features of the current frame. However, while these approaches primarily focus on enhancing in-frame regions, they often struggle to reconstruct critical out-of-frame areas near the sides of the ego-vehicle, although previous frames commonly contain valuable contextual information about these unseen regions. To address this limitation, we propose the Current-Centric Contextual 3D Fusion (C3DFusion) module, which generates hidden region-aware 3D feature geometry by explicitly aligning 3D-lifted point features from both current and historical frames. C3DFusion performs enhanced temporal fusion through two complementary techniques—historical context blurring and current-centric feature densification—which suppress noise from inaccurately warped historical point features by attenuating their scale, and enhance current point features by increasing their volumetric contribution. Simply integrated into standard SSC architectures, C3DFusion demonstrates strong effectiveness, significantly outperforming state-of-the-art methods on the SemanticKITTI and SSCBench-KITTI-360 datasets. Furthermore, it exhibits robust generalization, achieving notable performance gains when applied to other baseline models.
Transfer learning of diffusion models to smaller target domains is challenging, as naively fine-tuning the model often results in poor generalization. Test-time guidance methods help mitigate this by offering controllable improvements in image fidelity through a trade-off with sample diversity. However, this benefit comes at a high computational cost, typically requiring dual forward passes during sampling. We propose the Domain-guided Fine-tuning (DogFit) method, an effective guidance mechanism for diffusion transfer learning that maintains controllability without incurring additional computational overhead. DogFit injects a domain-aware guidance offset into the training loss, effectively internalizing the guided behavior during the fine-tuning process. The domain-aware design is motivated by our observation that during fine-tuning, the unconditional source model offers a stronger marginal estimate than the target model. To support efficient controllable fidelity–diversity trade-offs at inference, we encode the guidance strength value as an additional model input through a lightweight conditioning mechanism. We further investigate the optimal placement and timing of the guidance offset during training and propose two simple scheduling strategies, i.e., late-start and cut-off, which improve generation quality and training stability. Experiments on DiT and SiT backbones across six diverse target domains show that DogFit can outperform prior guidance methods in transfer learning in terms of FID and FD DINOV2 while requiring up to 2x fewer sampling TFLOPS.
Neural signed distance functions (SDFs) have been a vital representation to represent 3D shapes or scenes with neural networks. An SDF is an implicit function that can query signed distances at specific coordinates for recovering a 3D surface. Although implicit functions work well on a single shape or scene, they pose obstacles when analyzing multiple SDFs with high-fidelity geometry details, due to the limited information encoded in the latent space for SDFs and the loss of geometry details. To overcome these obstacles, we introduce a method to represent multiple SDFs in a common space, aiming to recover more high-fidelity geometry details with more compact latent representations. Our key idea is to take full advantage of the benefits of generalization-based and overfitting-based learning strategies, which manage to preserve high-fidelity geometry details with compact latent codes. Based on this framework, we also introduce a novel sampling strategy to sample training queries. The sampling can improve the training efficiency and eliminate artifacts caused by the influence of other SDFs. We report numerical and visual evaluations on widely used benchmarks to validate our designs and show advantages over the latest methods in terms of the representative ability and compactness.
RGB-based camouflaged object detection struggles in real-world scenarios where color and texture cues are ambiguous. While hyperspectral image offers a powerful alternative by capturing fine-grained spectral signatures, progress in hyperspectral camouflaged object detection (HCOD) has been critically hampered by the absence of a dedicated, large-scale benchmark. To spur innovation, we introduce HyperCOD, the first challenging benchmark for HCOD. Comprising 350 high-resolution hyperspectral images, It features complex real-world scenarios with minimal objects, intricate shapes, severe occlusions, and dynamic lighting to challenge current models.The advent of foundation models like the Segment Anything Model (SAM) presents a compelling opportunity. To adapt the Segment Anything Model (SAM) for HCOD, we propose HyperSpectral Camouflage-aware SAM (HSC-SAM). HSC-SAM ingeniously reformulates the hyperspectral image by decoupling it into a spatial map fed to SAM's image encoder and a spectral saliency map that serves as an adaptive prompt. This translation effectively bridges the modality gap. Extensive experiments show that HSC-SAM sets a new state-of-the-art on HyperCOD and generalizes robustly to other public HSI datasets. The HyperCOD dataset and our HSC-SAM baseline provide a robust foundation to foster future research in this emerging area.
Recent advancements in 3D Gaussian Splatting (3DGS) have demonstrated remarkable rendering quality, However, their substantial computational demands hinder practical deployment on resource-constrained devices. We propose a novel plug-and-play structured compression framework that significantly reduces computational overhead while maintaining rendering fidelity. We first discover that the statistical distribution of anchor vectors is decoupled from rendering quality. Based on this finding, we propose a distribution regularization method that enforces alignment to standard Gaussian distribution through KL divergence while optimizing Gaussian radius, significantly improving entropy coding efficiency. Second, we innovatively introduce an opacity-based probabilistic pruning mechanism that transforms pruning into an opacity optimization problem, achieving intelligent scene sparsification while allowing flexible adjustment according to hardware resources. Finally, we design a lightweight high-frequency compensation network that regards the high-frequency loss caused by over-compression as a residual and effectively recovers the high-frequency details lost during the compression process through residual learning. All modules are plug-and-play and can be seamlessly integrated into mainstream structured 3DGS frameworks. Extensive experiments on Synthetic-NeRF, Tanks&Temples, Mip-NeRF360 and DeepBlending datasets demonstrate that our method significantly reduces size by over 80x compared to vanilla 3DGS while simultaneously improving fidelity. Furthermore, it achieves a better size reduction and a 20% improvement in entropy encoding efficiency when compared to HAC, while meeting the requirements for real-time rendering.
LiDAR Semantic Scene Completion (SSC) in autonomous driving requires predicting both dense occupancy and semantic labels from sparse input point cloud. Existing methods typically adopt cascaded architecture for feature dilation and semantic abstraction, which blurs distinctive geometric patterns and reduces feature discriminability. Moreover, given an input, conventional processing of the ground truth labels overlooks voxel predictability in the target, resulting in ill-posed supervision and discards informative voxels. To address these limitations, we propose Sparse-Dense Net (SDNet), a dual-branch architecture that processes the input points through parallel sparse and dense encoders. The complementary features are aligned and fused using a Sparse Dense Feature Fusion (SDFF) module and further refined by a Feature Propagation (FP) module. Additionally, we introduce an input-aware label refinement strategy, including Sparse-Guided Filtering (SGF) to filter unpredictable targets and Ignored Voxel Recycling (IVR) to leverage informative ignored voxels for auxiliary supervision. These innovations enhance both feature learning and label quality. Extensive experiments on SemanticKITTI and nuScenes OpenOccupancy datasets validate the effectiveness of our approach, with SDNet achieving state-of-the-art performance on both datasets and ranking 1st on the official SemanticKITTI benchmark with 42.1 mIoU, outperforming the previous best by 4.2 (+11.1\%).
Mathematical Expression Recognition (MER) has made significant progress in recognizing simple expressions, but the robust recognition of complex mathematical expressions with many tokens and multiple lines remains a formidable challenge. In this paper, we first introduce CMER-Bench, a carefully constructed benchmark that categorizes expressions into three difficulty levels: easy, moderate, and complex. Leveraging CMER-Bench, we conduct a comprehensive evaluation of existing MER models and general-purpose multimodal large language models (MLLMs). The results reveal that while current methods perform well on easy and moderate expressions, their performance degrades significantly when handling complex mathematical expressions, mainly because existing public training datasets are primarily composed of simple samples. In response, we propose MER-17M and CMER-3M that are large-scale datasets emphasizing the recognition of complex mathematical expressions. The datasets provide rich and diverse samples to support the development of accurate and robust complex MER models. Furthermore, to address the challenges posed by the complicated spatial layout of complex expressions, we introduce a novel expression tokenizer, and a new representation called Structured Mathematical Language, which explicitly models the hierarchical and spatial structure of expressions beyond LaTeX format. Based on these, we propose a specialized model named CMERNet, built upon an encoder-decoder architecture and trained on CMER-3M. Experimental results show that CMERNet, with only 125 million parameters, significantly outperforms existing MER models and MLLMs on CMER-Bench.
Recent advances in instruction-based image editing have shown remarkable progress. However, existing methods remain limited to relatively simple editing operations, hindering real-world applications that require complex and compositional instructions. In this work, we address these limitations from the perspectives of architectural design, data, and evaluation protocols. Specifically, we identify two key challenges in current models: insufficient instruction compliance and background inconsistency. To this end, we propose MCIE-E1, a Multimodal Large Language Model–Driven Complex Instruction Image Editing method that integrates two key modules: a spatial-aware cross-attention module and a background-consistent cross-attention module. The former enhances instruction-following capability by explicitly aligning semantic instructions with spatial regions through spatial guidance during the denoising process, while the latter preserves features in unedited regions to maintain background consistency. To enable effective training, we construct a dedicated data pipeline to mitigate the scarcity of complex instruction-based image editing datasets, combining fine-grained automatic filtering via a powerful MLLM with rigorous human validation. Finally, to comprehensively evaluate complex instruction-based image editing, we introduce CIE-Bench, a new benchmark with two new evaluation metrics. Experimental results on CIE-Bench demonstrate that MCIE-E1 consistently outperforms previous state-of-theart methods in both quantitative and qualitative assessments, achieving a 23.96% improvement in instruction compliance.
Vision-Language Models (VLMs) have achieved impressive performance across various tasks, but often struggle to apply newly introduced visual concepts during inference. A common failure pattern is what we call Mixing Things Up: VLMs frequently confuse concept names, resulting in vague descriptions and failure to ground the concept correctly. Existing approaches mainly address person-related concepts through text prompts or tokenizer modifications. However, VLMs still miss or misinterpret untrained visual concepts, underscoring the need to learn new concepts directly from visual input, without relying on prior textual injection. To overcome these limitations, we propose BISCUIT (Basis-aligned Inference through Structured Concept Unification and Identification-aware Tuning), a two-step training method. Step I proposes a dual-stream structure-aware vision encoder that fuses RGB and edge-based embeddings within a shared basis space to enhance concept recognition. Step II enhances generation quality through identification-aware tuning, which encourages alignment between the generated text and the newly introduced visual concepts. Existing methods mainly focus on person concepts and lack comprehensive evaluation across diverse visual categories. We further propose a benchmark BiscuitVQA to evaluate VLMs performance on recognizing and applying novel image-introduced concepts across diverse concept types and task types, including real people, cartoons, animals, and symbolic content. We apply BISCUIT to LLaVA-1.5 and Qwen2.5-VL, achieving competitive results among open-source models and narrowing the gap to Gemini-2.5 and GPT-4o. Interestingly, our BISCUIT maintains strong generalization, showing minimal degradation on other downstream tasks.
Scene Text Editing (STE) aims to naturally modify text in images while preserving visual consistency, the decisive factors of which can be divided into three parts, i.e., text style, text content, and background. Previous methods have struggled with incomplete disentanglement of editable attributes, typically addressing only one aspect—such as editing text content—thus limiting controllability and visual consistency. To overcome these limitations, we propose TripleFDS, a novel framework for STE with disentangled modular attributes, and an accompanying dataset called SCB Synthesis. SCB Synthesis provides robust training data for triple feature disentanglement by utilizing the "SCB Group", a novel construct that combines three attributes per image to generate diverse, disentangled training groups. Leveraging this construct as a basic training unit, TripleFDS first disentangles triple features, ensuring semantic accuracy through inter-group contrastive regularization and preventing redundancy through intra-sample multi-feature orthogonality. In the synthesis phase, TripleFDS performs feature remapping to prevent "shortcut" phenomena during reconstruction and mitigate potential feature leakage. Trained on 125,000 SCB Groups, TripleFDS achieves state-of-the-art image fidelity (SSIM of 44.54) and text accuracy (ACC of 93.58%) on the mainstream STE benchmarks. Besides superior performance, the more flexible editing of TripleFDS supports new operations such as style replacement and background transfer.
3D Gaussian Splatting (3DGS) has recently demonstrated significant potential for streaming dynamic scenes, enabling the synthesis of photo-realistic and real-time free-viewpoint videos (FVVs). Conventional streaming pipelines optimize each frame independently, i.e., the attribute of the 3D Gaussians (3DGs) responsible for the static regions are supposed to be identical across all frames but are changed in the optimization process, thus causing temporal color inconsistency and visual flickering artifacts in the static regions. To tackle this, we propose CPOStream, which utilizes a prediction and observation module to determine the state of 3DG. Specifically, the prediction module records those 3DGs that are inactive in the past K frames and those would be ignored in the optimization process of the current frame reconstruction. Thus, the attributes of those 3DGs would be kept consistent across the past K frames, guaranteeing the temporal consistence. Additionally, the observation module conducts motion detection, and recognizes those new 3DGs which are not recorded in the prediction module and are first detected by the observation module in the past K frames. The attributes of those 3DGs are optimized during the current frame reconstruction. Experiments on multiple real-world FVV benchmarks show that CPOStream substantially reduces temporal flickering and improves reconstruction fidelity, achieving state‑of‑the‑art performance.
Prompt-driven scene synthesis allows users to generate complete 3D environments from textual descriptions. Current text-to-scene methods often struggle with complex geometries and object transformations, and tend to show weak adherence to complex instructions. We address these limitations by introducing Reason-3D, a text-to-scene model powered by large reasoning models (LRMs). Reason-3D integrates object retrieval using captions covering physical, functional, and contextual attributes. Reason-3D then places the selected objects based on implicit and explicit layout constraints, and refines their positions with collision-aware spatial reasoning. Evaluated on instructions ranging from simple to complex indoor configurations, Reason-3D significantly outperforms previous methods in human-rated visual fidelity, adherence to constraints, and asset retrieval quality. Beyond its contribution to the field of text-to-scene generation, our work showcases the advanced spatial reasoning abilities of modern LRMs. Additionally, we release the codebase to further the research in object retrieval and placement with LRMs.
In recent years, Contrastive Language-Image Pretraining (CLIP) has been widely applied to Weakly Supervised Semantic Segmentation (WSSS) tasks due to its powerful cross-modal semantic understanding capabilities. This paper proposes a novel Semantic and Spatial Rectification (SSR) method to address the limitations of existing CLIP-based weakly supervised semantic segmentation approaches: over-activation in non-target foreground regions and background areas. Specifically, at the semantic level, the Cross-Modal Prototype Alignment (CMPA) establishes a contrastive learning mechanism to enforce feature space alignment across modalities, reducing inter-class overlap while enhancing semantic correlations, to rectify over-activation in non-target foreground regions effectively; at the spatial level, the Superpixel-Guided Correction (SGC) leverages superpixel-based spatial priors to precisely filter out interference from non-target regions during affinity propagation, significantly rectifying background over-activation. Extensive experiments on the PASCAL VOC and MS COCO datasets demonstrate that our method outperforms all single-stage approaches, as well as more complex multi-stage approaches, achieving mIoU scores of 79.5% and 50.6%, respectively.
The potential of Generalized Category Discovery (GCD) lies in its ability to identify previously undiscovered patterns in both labeled and unlabeled data by leveraging insights from partially labeled training samples. However, interference can arise due to the model's dual focus on discovering both novel and known categories, often leading to conflicts that obscure true patterns in the dataset. This paper presents a divide-and-conquer framework, Foundation-Adaptive Integrated Refinement (FAIR), which fine-tunes pretrained foundational weights for various purposes, divided into Foundation (pretrained weights), Adaptive (weights fine-tuned with a variance-preserving loss), and Integrated (weights adjusted for both labeled and unlabeled data). The Adaptive utilizes a newly proposed adaptive contrastive loss that introduces variances within classes to preserve the individuality of representations. The Integrated addresses inherent estimation errors while dynamically estimating the number of categories, incorporating a cosine-based perturbation mechanism as a relaxed margin to accommodate potential ground-truth deviations, rather than relying on biased estimates. Extensive experiments on six benchmark datasets demonstrate our method's effectiveness, outperforming state-of-the-art algorithms, especially on fine-grained datasets.
Existing cross-modal pedestrian detection (CMPD) employs complementary information from RGB and thermal-infrared (TIR) modalities to detect pedestrians in 24h-surveillance systems. RGB captures rich pedestrian details under daylight, while TIR excels at night. However, TIR focuses primarily on the person's silhouette, neglecting critical texture details essential for detection. While the near-infrared (NIR) captures texture under low-light conditions, which effectively alleviates performance issues of RGB and detail loss in TIR, thereby reducing missed detections. To this end, we construct a new Triplet RGB–NIR–TIR (TRNT) dataset, comprising 8,281 pixel-aligned image triplets, establishing a comprehensive foundation for algorithmic research. However, due to the variable nature of real-world scenarios, imaging devices may not always capture all three modalities simultaneously. This results in input data with unpredictable combinations of modal types, which challenge existing CMPD methods that fail to extract robust pedestrian information under arbitrary input combinations, leading to significant performance degradation. To address these challenges, we propose the Adaptive Uncertainty-aware Network (AUNet) for accurately discriminating modal availability and fully utilizing the available information under uncertain inputs. Specifically, we introduce Unified Modality Validation Refinement (UMVR), which includes an uncertainty-aware router to validate modal availability and a semantic refinement to ensure the reliability of information within the modality. Furthermore, we design a Modality-Aware Interaction (MAI) module to adaptively activate or deactivate its internal interaction mechanisms per UMVR output, enabling effective complementary information fusion from available modalities. AUNet enables accurate modality validation and robust inference without fixed modality pairings, facilitating the effective fusion of RGB, NIR, and TIR information across diverse inputs.
Large-scale vision-language models (VLMs) embedded with expansive representations and visual concepts have showcased significant potential in image and text understanding. Efficiently adapting VLMs such as CLIP to downstream tasks like few-shot image classification has garnered growing attention, with prompt learning emerging as a representative approach. However, most existing prompt-based adaptation methods, which rely solely on coarse-grained textual prompts, suffer from limited performance and interpretability when handling domain tasks that require specific knowledge. This results in a failure to satisfy the stringent trustworthiness requirements of Explainable Artificial Intelligence (XAI) in high-risk scenarios like healthcare. To address this issue, we propose a Knowledge-Enhanced Explainable Prompting (KEEP) framework that leverages fine-grained domain-specific knowledge to enhance the adaptation process of VLMs across various domains and image modalities. By incorporating retrieval augmented generation and domain foundation models, our framework can provide more reliable image-wise knowledge for prompt learning in various domains, alleviating the lack of fine-grained annotations, while offering both visual and textual explanations. Extensive experiments and explainability analyses conducted on eight datasets of different domains and image modalities demonstrate that our method simultaneously achieves superior performance and interpretability, highlighting the effectiveness of the collaboration between foundation models and XAI.
We present MoBGS, a novel motion deblurring 3D Gaussian Splatting (3DGS) framework capable of reconstructing sharp and high-quality novel spatio-temporal views from blurry monocular videos in an end-to-end manner. Existing dynamic novel view synthesis (NVS) methods are highly sensitive to motion blur in casually captured videos, resulting in significant degradation of rendering quality. While recent approaches address motion-blurred inputs for NVS, they primarily focus on static scene reconstruction and lack dedicated motion modeling for dynamic objects. To overcome these limitations, our MoBGS introduces a novel Blur-adaptive Latent Camera Estimation (BLCE) method using a proposed Blur-adaptive Neural Ordinary Differential Equation (ODE) solver for effective latent camera trajectory estimation, improving global camera motion deblurring. In addition, we propose a Latent Camera-induced Exposure Estimation (LCEE) method to ensure consistent deblurring of both a global camera and local object motions. Extensive experiments on the Stereo Blur dataset and real-world blurry videos show that our MoBGS significantly outperforms the very recent methods, achieving state-of-the-art performance for dynamic NVS under motion blur.
Recent progress in text-to-image (T2I) generative models has led to significant improvements in generating high-quality images aligned with text prompts. However, these models still struggle with prompts involving multiple objects, often resulting in object neglect or object mixing. Through extensive studies, we identify four problematic scenarios, Similar Shapes, Similar Textures, Dissimilar Background Biases, and Many Objects, where inter-object relationships frequently lead to such failures. Motivated by two key observations about CLIP embeddings, we propose DOS (Directional Object Separation), a method that modifies three types of CLIP text embeddings before passing them into text-to-image models. Experimental results show that DOS consistently improves the success rate of multi-object image generation and reduces object mixing. In human evaluations, DOS significantly outperforms four competing methods, receiving 26.24%-43.04% more votes across four benchmarks. These results highlight DOS as a practical and effective solution for improving multi-object image generation.
Flow Matching has emerged as a powerful framework for learning continuous transformations between distributions, enabling high-fidelity generative modeling. This work introduces Symmetrical Flow Matching (SymmFlow), a new formulation that unifies semantic segmentation, classification, and image generation within a single model. Using a symmetric learning objective, SymmFlow models forward and reverse transformations jointly, ensuring bi-directional consistency, while preserving sufficient entropy for generative diversity. A new training objective is introduced to explicitly retain semantic information across flows, featuring efficient sampling while preserving semantic structure, allowing for one-step segmentation and classification without iterative refinement. Unlike previous approaches that impose strict one-to-one mapping between masks and images, SymmFlow generalizes to flexible conditioning, supporting both pixel-level and image-level class labels. Experimental results on various benchmarks demonstrate that SymmFlow achieves state-of-the-art performance on semantic image synthesis, obtaining FID scores of 11.9 on CelebAMask-HQ and 7.0 on COCO-Stuff with only 25 inference steps. Additionally, it delivers competitive results on semantic segmentation and shows promising capabilities in classification tasks.
Recent advancements in text-to-image (T2I) generation have led to the emergence of highly expressive models such as diffusion transformers (DiTs), exemplified by FLUX. However, their massive parameter sizes lead to slow inference, high memory usage, and poor deployability. Existing acceleration methods (e.g., single-step distillation and attention pruning) often suffer from significant performance degradation and incur substantial training costs. To address these limitations, we propose FastFLUX, an architecture-level pruning framework designed to enhance the inference efficiency of FLUX. At its core is the Block-wise Replacement with Linear Layers (BRLL) method, which replaces structurally complex residual branches in ResBlocks with lightweight linear layers while preserving the original shortcut connections for stability. Furthermore, we introduce Sandwich Training (ST), a localized fine-tuning strategy that leverages LoRA to supervise neighboring blocks, mitigating performance drops caused by structural replacement. Experiments show that our FastFLUX maintains high image quality under both qualitative and quantitative evaluations, while significantly improving inference speed, even with 20% of the hierarchy pruned.