WACV 2026 论文列表

Identity-preserving portrait generation has achieved tremendous advancements with the development of diffusion models. However, multi-ID generation remains challenging due to degraded identity fidelity and insufficient control over layout, pose, and expression. To address these challenges, we propose GroupPortrait, a novel approach for multi-ID portrait generation with three key innovations:(1) LatentID for high-fidelity identity preservation, (2) Facial Controller enabling layout guidance and fine-grained facial control, and (3) Mask-Attention Controller allocating identity embeddings to specific facial regions. First, the LatentID module improves identity preservation by adding LatentID loss during training. It maps latent representations to identity features and uses ID consistency loss for feedback training to improve identity retention. Since LatentID loss is calculated in latent space, it is more efficient in terms of time and GPU usage compared to the method that calculates ID loss in pixel space. Second, to enhance layout and facial controllability, the Facial Controller utilizes 3D Morphable Models (3DMM) to acquire facial shapes, poses, and expressions for each individual, imposing strong spatial conditions during the diffusion process. Finally, we propose a novel Mask-Attention Controller for multi-ID generation, which distributes ID embeddings into target facial regions by aligning the cross-attention map of LatentID with the given facial region masks. Extensive experiments demonstrate that GroupPortrait can generate group portraits with high fidelity, local harmony, and controllability.

Object detection with 3D radar is essential for 360^ \circ automotive perception, but radar's long wavelengths produce sparse and irregular reflections that challenge traditional grid and sequence-based convolutional and transformer detectors. This paper introduces Graph Query Networks (GQN), an attention-based framework that models objects sensed by radar as graphs, to extract individualized relational and contextual features. GQN employs a novel concept of graph queries to dynamically attend over the bird's-eye view (BEV) space, constructing object-specific graphs processed by two novel modules: EdgeFocus for relational reasoning and DeepContext Pooling for contextual aggregation. On the NuScenes dataset, GQN improves relative mAP by up to +53%, including a +8.2% gain over the strongest prior radar method, while reducing peak graph construction overhead by 80% with moderate FLOPs cost.

Recently, significant efforts were made towards Variational Autoencoder (VAE) -based prototypical Self Explainable Models (SEM) for image classification. The princi-ple is to learn class-specific prototypes that can be projected back into the image spacethanks to the decoding branch of a VAE. However, existing VAE-based SEM fail to rep-resent properly the distribution of training samples in the embedding space, requiringto define specific additional constraints as diversity or orthogonality. In this work, wepropose to define the prototypes as the components of a Gaussian Mixture VAE (GM-VAE) that is an approximation of the distribution of training samples. We show that thisdefinition allows to produce relevant and diverse prototypes providing a probabilistic ex-planation of the model without assigning prototypes to a specific class. We support ourdefinition with extensive experimentation and comparison with previous self-explainableapproaches.

Video-based gaze estimation methods aim to capture the inherently temporal dynamics of human eye gaze from multiple image frames. However, since models must capture both spatial and temporal relationships, performance is limited by the feature representations within a frame but also between multiple frames.We propose Spatio-Temporal Gaze Network (ST-Gaze), a model that combines an existing CNN backbone with dedicated channel attention and self-attention modules to fuse eye and face features optimally. The fused features are then treated as a spatial sequence, allowing for the capture of an intra-frame context, which is then propagated through time to model inter-frame dynamics. We evaluated our method on the EVE dataset and show that ST-Gaze achieves state-of-the-art performance both with and without person-specific adaptation. Additionally, our ablation study provides further insights into the model performance, showing that preserving and modelling intra-frame spatial context with our spatio-temporal recurrence is fundamentally superior to premature spatial pooling. As such, our results pave the way towards robust video-based gaze estimation.

Recent state-of-the-art (SOTA) methods in 3D human pose estimation (HPE) typically prioritize lifting 2D pose coordinates to 3D but tend to underemphasize the importance of generalizing under real-world conditions with noisy 2D inputs from off-the-shelf 2D detector. In this paper, we introduce Graph Attention Fusion Network (GAtFuN), a novel motion-aware framework that integrates our spatial and temporal graph attention mechanisms to explicitly model joint velocities and motion transformations, resulting in more stable and coherent 3D pose predictions despite being trained with the same dataset pipeline as other SOTA methods. GAtFuN achieves a 7.8% improvement in MPJPE over the current SOTA on the Human3.6M dataset and a 1.9% improvement on the MPI-INF-3DHP dataset, while demonstrating more robust performance on the 3DPW dataset in the wild.

Recent advances in diffusion models have significantly improved image and video synthesis. In addition, several concept control methods have been proposed to enable fine-grained, continuous, and flexible control over free-form text prompts. However, these methods not only require intensive training time and GPU memory usage to learn the sliders or embeddings but also need to be retrained for different diffusion backbones, limiting their scalability and adaptability. To address these limitations, we introduce Text Slider, a lightweight, efficient and plug-and-play framework that identifies low-rank directions within a pre-trained text encoder, enabling continuous control of visual concepts while significantly reducing training time, GPU memory consumption, and the number of trainable parameters. Furthermore, Text Slider supports multi-concept composition and continuous control, enabling fine-grained and flexible manipulation in both image and video synthesis. We show that Text Slider enables smooth and continuous modulation of specific attributes while preserving the original spatial layout and structure of the input. Text Slider achieves significantly better efficiency: 5xfaster training than Concept Slider and 47xfaster than Attribute Control, while reducing GPU memory usage by nearly 2xand 4x, respectively.

Lifelog images captured automatically by wearable cameras serve as effective cues that induce Autobiographic Memory Recall (AMR) of social interactions. This is very useful for personalized memory interventions. However, manual selection of images for such therapy imposes significant load on the caregivers who need to browse through a voluminous collection of images. To reduce this load, automated tools that identify moments involving significant engagement of the camera wearer in social interactions are needed. To achieve this, we re-annotate images extracted from public lifelog datasets for the presence of non-verbal social signals and the perceived engagement of the life-logger during interactions. We use this data to develop models and explore how social signals and the detected intensity of social engagement are helpful for predicting AMR. We show that understanding visual social engagement can enhance AMR prediction, demonstrating the potential of the models in reducing caregivers' effort.

Pre-trained text-to-image (T2I) diffusion models produce high-quality images from random noise guided by textual prompts, but classifier-free guidance (CFG) often yields suboptimal samples. Retraining diffusion models can enhance performance, though the associated computational cost is often prohibitive. To address this, researchers have proposed modifying the CFG step through perturbation techniques for cost-effective sampling improvements. For instance, Smoothed Energy Guidance (SEG) enhances image quality by reducing the curvature of the energy landscape in self-attention layers. However, SEG's indiscriminate curvature reduction can suppress fine details and cause prompt misalignment. We introduce Orthogonal Smoothed Energy Guidance (OSEG), which selectively smooths the orthogonal components of token embeddings to preserve critical details while maintaining prompt alignment. Theoretical analysis justifies OSEG's effectiveness, supported by empirical comparisons with recent methods. By plugging into text-to-video (T2V) models, OSEG consistently improves visual and temporal coherence. Finally, we present extensive metric comparisons to demonstrate the efficiency of OSEG for both image and video domains. Project homepage: https://osegguidance. github.io/

In recent years, personalization, which utilizes user-specific data to generate tailored responses, has been increasingly adopted in user-centric domains. However, while Large Language Models (LLMs) are actively researched, the exploration of the personalization capabilities of Large Vision-Language Models (LVLMs) remains limited. To systematically evaluate the personalization ability of LVLMs, we introduce PerVL-Bench, a synthetic benchmark specifically designed for this purpose. PerVL-Bench incorporates user-specific data, including multiple images and long text information, and provides two types of QA pairs. Furthermore, we use PerVL-Bench to comprehensively evaluate the essential capabilities for personalization in current state-of-the-art LVLMs. Through this evaluation, we reveal the limitations of current models in multimodal personalization and provide insights for the development of personalized LVLMs. We publicly release PerVL-Bench and the accompanying experimental code to advance future research: https://github.com/MSungK/PerVL-Bench

Accurate segmentation of colorectal polyps plays a vital role in the early diagnosis and prevention of colorectal cancer. Despite notable progress, existing methods struggle with limited region adaptability due to fixed receptive fields, lack explicit boundary modeling, and are prone to interference from background noise, leading to suboptimal segmentation results. To address these issues, we propose FARF-Net, a novel edge-aware segmentation framework that leverages frequency-domain adaptive receptive fields. Built upon the Pyramid Vision Transformer v2 backbone, FARF-Net introduces three tailored components. Specifically, EdgeKAN module applies Kolmogorov-Arnold Networks for channel-wise nonlinear modeling, enhancing local edge semantics and boundary detail representation. Adaptive Receptive Field module adjusts spatial receptive fields based on localized frequency energy, boosting sensitivity to high-frequency boundaries. Frequency-Guided Dual-Supervision Decoder integrates high-frequency structural features and boundary priors to refine edge predictions and suppress irrelevant high-frequency background noise. Extensive experiments on five public polyp segmentation benchmarks demonstrate that FARF-Net consistently surpasses state-of-the-art methods. Notably, it achieves superior boundary reconstruction and robustness in challenging cases such as blurred contours and small polyps.

In most multi-modal weakly supervised video anomaly detection (WSVAD) methods, cross-modal alignment relies on cosine similarity measure, which captures only directional consistency but neglects feature magnitude information. Moreover, the existing multi-instance learning frameworks usually adopt the Top-k selection strategy, which is difficult to adapt to the anomalous proportions of different videos, resulting in missed detection of anomalous segments and introduction of label noise. To address these problems, an alignment and selection enhanced WSVAD (ASE-WSVAD) method is proposed. ASE-WSVAD combines cross-modal alignment method based on the fused similarity measure (CA-FSM) with dual-scale adaptive selection (DSAS) to improve semantic consistency and detection performance. Specifically, visual-textual alignment is implemented by a fusion of complementary similarity measures (i.e., cosine and Euclidean distance similarity measures) so that the alignment objective jointly leverages both direction and magnitude. DSAS combines local instance-level selection (LILS) with global batch-aware selection (GBAS) to efficiently detect anomalous segments and handle videos with varying proportions of anomalies. Experimental results demonstrate that ASE-WSVAD achieves the state-of-the-art performance with AUC value of 89.00% and AP value of 85.44% on UCF-Crime and XD-Violence, respectively.

Flood prediction is critical for emergency planning and response to mitigate human and economic losses. Traditional physics-based hydrodynamic models generate high-resolution flood maps using numerical methods requiring fine-grid discretization; which are computationally intensive and impractical for real-time large-scale applications. While recent studies have applied convolutional neural networks for flood map super-resolution with good accuracy and speed, they suffer from limited generalizability to unseen areas. In this paper, we propose a novel approach that leverages latent diffusion models to perform super-resolution on coarse-grid flood maps, with the objective of achieving the accuracy of fine-grid flood maps while significantly reducing inference time. Experimental results demonstrate that latent diffusion models substantially decrease the computational time required to produce high-fidelity flood maps without compromising on accuracy, enabling their use in real-time flood risk management. Moreover, diffusion models exhibit superior generalizability across different physical locations, with transfer learning further accelerating adaptation to new geographic regions. Our approach also incorporates physics-informed inputs, addressing the common limitation of black-box behavior in machine learning, thereby enhancing interpretability. Code is available at https://github.com/neosunhan/flood-diff.

Embodied Question Answering (EQA) combines visual scene understanding, goal-directed exploration, spatial and temporal reasoning under partial observability. A central challenge is to confine physical search to question-relevant subspaces while maintaining a compact, actionable memory of observations. Furthermore, for real-world deployment, fast inference time during exploration is crucial. We introduce FAST-EQA, a question-conditioned framework that (i) identifies likely visual targets, (ii) scores global regions of interest to guide navigation, and (iii) employs Chain-of-Thought (CoT) reasoning over visual memory to answer confidently. FAST-EQA maintains a bounded scene memory that stores a fixed-capacity set of region-target hypotheses and updates them online, enabling robust handling of both single- and multi-target questions without unbounded growth. To expand coverage efficiently, a global exploration policy treats narrow openings and doors as high-value frontiers, complementing local target seeking with minimal computation. Together, these components focus the agent's attention, improves scene coverage, and improve answer reliability while running substantially faster than prior approaches. On HMEQA and EXPRESS-Bench, FAST-EQA achieves state-of-the-art performance, while performing competitively on OpenEQA and MT-HM3D.

We introduce WarpRF, a training-free general-purpose framework for quantifying the uncertainty of radiance fields. Built upon the assumption that photometric and geometric consistency should hold among images rendered by an accurate model, WarpRF quantifies its underlying uncertainty from an unseen point of view by leveraging backward warping across viewpoints, projecting reliable renderings to the unseen viewpoint and measuring the consistency with images rendered there. WarpRF is simple and inexpensive, does not require any training, and can be applied to any radiance field implementation for free. WarpRF excels at both uncertainty quantification and downstream tasks, e.g., active view selection and active mapping, outperforming any existing method tailored to specific frameworks.

What role does a single vision token play inside a multimodal large language model (MLLM)? Despite recent successes, most MLLMs adopt a simple design: a projected visual feature z = P(f_x) prepended to the text sequence. Yet it remains unclear whether this vector merely provides context or actively steers generation. We propose a geometric probing framework that analyzes latent-token alignment, intrinsic dimensionality, and perturbation sensitivity. Across four datasets and three representative MLLMs (LLaVA, BLIP-2, Kosmos-2), we find clear operating regimes: BLIP-2 enforces rigid low-rank compression with strong alignment but near-zero sensitivity, LLaVA exhibits flexible high-dimensional mappings with high responsiveness, and Kosmos-2 balances between them. These signatures correlate with downstream behavior--SQA correctness and VQAv2 hallucination severity--showing that reduced or excessive sensitivity predicts unreliable grounding. Our results highlight geometry as a diagnostic lens for vision-language coupling and offer actionable guidance for projection design, alignment objectives, and user-steerable multimodal generation.

Self-supervised pre-training based on next-token prediction has enabled large language models to capture the underlying structure of text, and has led to unprecedented performance on a large array of tasks when applied at scale. Similarly, autonomous driving generates vast amounts of spatiotemporal data, alluding to the possibility of harnessing scale to learn the underlying geometric and semantic structure of the environment and its evolution over time. In this direction, we propose a geometric and semantic self-supervised pre-training method, GASP, that learns a unified representation by predicting, at any queried future point in spacetime, (1) general occupancy, capturing the evolving structure of the 3D scene; (2) ego occupancy, modeling the ego vehicle path through the environment; and (3) distilled high-level features from a vision foundation model. By modeling geometric and semantic 4D occupancy fields instead of raw sensor measurements, the model learns a structured, generalizable representation of the environment and its evolution through time. We validate GASP on multiple autonomous driving benchmarks, demonstrating significant improvements in semantic occupancy forecasting, online mapping, and ego trajectory prediction. Our results demonstrate that continuous 4D geometric and semantic occupancy prediction provides a scalable and effective pre-training paradigm for autonomous driving. Code will be released.

Accurate swimming performance monitoring has traditionally relied on wearable sensors, which can disrupt natural technique and are often impractical in competitive settings. In this paper, we present a fully vision-based system for automatic swimmer analysis using overhead drone footage, removing the need for any body-mounted device or underwater equipment. By fine-tuning pose estimation models for aerial aquatic conditions, our approach robustly extracts full-body swimmer skeletons even under challenging scenarios such as splashes and partial occlusions. From these poses, we classify swimming strokes, compute instantaneous speed, estimate lap times, and count individual strokes. Unlike existing methods, our system provides scalable, unobtrusive, and infrastructure-free tracking. Evaluated on real-world drone-captured swimming competition data, our method achieves a median speed estimation error below 4% (under 0.05 m/s), a median lap time error of just 0.03s, and stroke count errors typically under one stroke per lap.

Implicit Neural Representations (INRs) model signals as continuous, differentiable functions.However, monolithic INRs scale poorly with data dimensionality, leading to excessive training costs.We propose F-INR, a framework that addresses this limitation by factorizing a high-dimensional INR into a set of compact, axis-specific sub-networks based on functional tensor decomposition.These sub-networks learn low-dimensional functional components that are then combined via tensor operations.This factorization reduces computational complexity while additionally improving representational capacity.F-INR is both architecture- and decomposition-agnostic.It integrates with various existing INR backbones (e.g., SIREN, WIRE, FINER, Factor Fields) and tensor formats (e.g., CP, TT, Tucker), offering fine-grained control over the speed-accuracy trade-off via the tensor rank and mode.Our experiments show F-INR accelerates training by up to 20xand improves fidelity by over 6.0 dB PSNR compared to state-of-the-art INRs.We validate these gains on diverse tasks, including image representation, 3D geometry reconstruction, and neural radiance fields.We further show F-INR's applicability to scientific computing by modeling complex physics simulations. Thus, F-INR provides a scalable, flexible, and efficient framework for high-dimensional signal modeling.

Visual Planning for Assistance (VPA) aims to predict a sequence of user actions required to achieve a specified goal based on a video showing the user's progress. Although recent advances in multimodal large language models (MLLMs) have shown promising results in video understanding, long-horizon visual planning remains a challenging problem. We identify two challenges in training large MLLMs for video-based planning tasks: (1) scarcity of procedural annotations, limiting the model's ability to learn procedural task dynamics effectively, and (2) inefficiency of next-token prediction objective to explicitly capture the structured action space for visual planning when compared to free-form, natural language. To tackle data scarcity, we introduce Auxiliary Task Augmentation. We design and train our model on auxiliary tasks relevant to long-horizon video-based planning (e.g., goal prediction) to augment the model's planning ability. To more explicitly model the structured action space unique to visual planning tasks, we leverage Multi-token Prediction, extending traditional next-token prediction by using multiple heads to predict multiple future tokens during training. Our approach, VideoPlan, achieves state-of-the-art VPA performance on the COIN and CrossTask datasets, surpassing prior methods by 7.3% and 3.4%, respectively, when predicting 3 future actions. We further extend our method to the challenging Ego4D Long-term Action Anticipation task, and show that it is on par with the state-of-the-art approaches despite not using specialized egocentric features. Code is available at https://github.com/CeeZh/VideoPlan.