WACV 2026 Papers

Volumetric learning underpins many 3D vision tasks such as completion, reconstruction, and mesh generation, yet training objectives still rely on Chamfer Distance (CD) or Earth Mover's Distance (EMD), which fail to balance recall and precision. We propose Quality-Aware Loss (QAL), a drop-in replacement for CD/EMD that combines a coverage-weighted nearest-neighbor term with an uncovered-ground-truth attraction term, explicitly decoupling recall and precision into tunable components. Across diverse pipelines, QAL achieves consistent coverage gains, improving by an average of +4.3 pts over CD and +2.8 pts over the best alternatives. These improvements reliably recover thin structures and under-represented regions that CD/EMD overlook. Extensive ablations confirm stable performance across hyperparameters and output resolutions, while full retraining on PCN and ShapeNet demonstrates generalization across datasets and backbones. Moreover, QAL-trained completions yield higher grasp scores under GraspNet evaluation, showing that improved coverage translates directly into more reliable robotic manipulation. QAL thus offers a principled, interpretable, and practical objective for robust 3D vision and safety-critical robotics pipelines.

One-shot styled handwriting image generation, despite achieving impressive results in recent years, remains challenging due to the difficulty in capturing the intricate and diverse characteristics of human handwriting by using solely a single reference image. Existing methods still struggle to generate visually appealing and realistic handwritten images and adapt to complex, unseen writer styles, struggling to isolate invariant style features (e.g., slant, stroke width, curvature) while ignoring irrelevant noise. To tackle this problem, we introduce Patch Contrastive Enhancement and Style-Aware Quantization via Denoising Diffusion (CONSTANT), a novel one-shot handwriting generation via diffusion model. CONSTANT leverages three key innovations: 1) a Style-Aware Quantization (SAQ) module that models style as discrete visual tokens capturing distinct concepts; 2) a contrastive objective to ensure these tokens are well-separated and meaningful in the embedding style space; 3) a latent patch-based contrastive (LLatentPCE) objective help improving quality and local structures by aligning multiscale spatial patches of generated and real features in latent space. Extensive experiments and analysis on benchmark datasets from multiple languages, including English, Chinese, and our proposed ViHTGen dataset for Vietnamese, demonstrate the superiority of adapting to new reference styles and producing highly detailed images of our method over state-of-the-art approaches. Code is available at GitHub https://github.com/duylebkHCM/CONSTANT

Large vision-language models (VLMs) enable intuitive visual search using natural language queries. However, improving their performance often requires fine-tuning and scaling to larger model variants. In this work, we propose a mechanism inspired by traditional text-based search to improve retrieval performance at inference time: relevance feedback. While relevance feedback can serve as an alternative to fine-tuning, its model-agnostic design also enables use with fine-tuned VLMs. Specifically, we introduce and evaluate four feedback strategies for VLM-based retrieval. First, we revise classical pseudo-relevance feedback (PRF), which refines query embeddings based on top-ranked results. To address its limitations, we propose generative relevance feedback (GRF), which uses synthetic captions for query refinement. Furthermore, we introduce an attentive feedback summarizer (AFS), a custom transformer-based model that integrates multimodal fine-grained features from relevant items. Finally, we simulate explicit feedback using ground-truth captions as an upper-bound baseline. Experiments on Flickr30k and COCO with three VLM backbones show that GRF, AFS, and explicit feedback improve retrieval performance by 3-5% in MRR@5 for smaller VLMs, and 1-3% for larger ones, compared to retrieval with no feedback. Moreover, AFS, similarly to explicit feedback, mitigates query drift and is more robust than GRF in iterative, multi-turn retrieval settings. Our findings demonstrate that relevance feedback can consistently enhance retrieval across VLMs and open up opportunities for interactive and adaptive visual search.

Backdoor attacks are emerging threats to deep neural networks, which typically embed malicious behaviors into a victim model by injecting poisoned samples. Adversaries can activate the injected backdoor during inference by presenting the trigger on input images. Prior defensive methods have achieved remarkable success in countering dirty-label backdoor attacks where the labels of poisoned samples are often mislabeled. However, these approaches do not work for a recent new type of backdoor -- clean-label backdoor attacks that imperceptibly modify poisoned data and hold consistent labels. More complex and powerful algorithms are demanded to defend against such stealthy attacks. In this paper, we propose UltraClean, a general framework that simplifies the identification of poisoned samples and defends against both dirty-label and clean-label backdoor attacks. Given the fact that backdoor triggers introduce adversarial noise that intensifies in feed-forward propagation, UltraClean first generates two variants of training samples using off-the-shelf denoising functions. It then measures the susceptibility of training samples leveraging the error amplification effect in DNNs, which dilates the noise difference between the original image and denoised variants. Lastly, it filters out poisoned samples based on the susceptibility to thwart the backdoor implantation. Despite its simplicity, UltraClean achieves a superior detection rate across various datasets and significantly reduces the backdoor attack success rate while maintaining a decent model accuracy on clean data, outperforming existing defensive methods by a large margin. Code is available at https://github.com/bxz9200/UltraClean.

Text-conditioned diffusion models have revolutionized the field of controllable real image editing, enabling high-fidelity and precise image manipulation. Recent methods target specific editing tasks, using internal representations from reconstruction to ensure consistency. Although effective for single tasks, they fail to balance precision and consistency across diverse image editing tasks. In this work, we propose a novel inference-time real-image editing framework that enables executing multiple editing tasks by tuning editing operators. Our key insight is to treat real image editing as a multi-objective optimization problem, optimizing editing operators for a Pareto optimal solution that balances editing accuracy and consistency at each denoising iteration. Additionally, we design a benchmark for operator-guided real-image editing that covers various local and global editing tasks. Extensive experimental evaluations demonstrate the method's effectiveness in executing precise edits while preserving image fidelity across all tasks, thereby establishing it as the new state-of-the-art.

Collaborative inference (CI) is hampered by high communication costs and privacy risks, with existing defenses often forcing a trade-off between efficiency and formal privacy guarantees. In this work, we present a framework that leverages activation sparsity as a dual-purpose mechanism to address both challenges simultaneously. Our approach uses a lightweight Sparse Autoencoder (SAE) to learn a sparse representation, which is then protected by a novel two-channel noise mechanism grounded in information theory. This design provides a tunable privacy budget while remaining computationally inexpensive. Evaluations on CIFAR-10, Tiny-ImageNet, and FaceScrub show that our method achieves a state-of-the-art privacy-utility trade-off, sustaining high accuracy at sparsity levels of up to 97%, while offering superior resilience against strong model inversion attacks. Our results underline that sparsity can be transformed from an effective compression tool into a powerful and theoretically-grounded privacy defense, paving the way for more practical and trustworthy CI systems. We provide code at https://github.com/an7123/privacy_ci.

Unsupervised domain adaptation for object detection (UDAOD) aims to adapt the source detector to the target domain using labeled source data and unlabeled target data. To mitigate the gap between the two domains, existing methods employ the Mean Teacher (MT) framework for domain adaptation, selecting high-quality pseudo-labels generated by the teacher model to supervise the training of the student model. However, the pseudo-labels generated by the teacher model often contain a high proportion of false positive labels, which can mislead the student model and result in a decline in overall performance. In this paper, we propose a novel data augmentation strategy for domain adaptive object detection, Crop Adaptive Teacher (CropAT), to address this problem and improve model performance. We leverage prompt tuning on an off-the-shelf image editing model to generate target-like images from source data, aiming to reduce the domain gap. Additionally, we insert object crops from these target-like images into the unlabeled target data to increase the correct labels within pseudo-labels, consequently decreasing the proportion of false positive pseudo-labels. Our method outperforms existing approaches across multiple benchmarks. For Cityscapes (source) to Foggy Cityscapes (target) adaptation, CropAT achieves 53.2% mAP on the target domain, surpassing the baseline method and the previous state-of-the-art (SOTA) by 3.9% and 0.7%. For PASCAL VOC (source) to Clipart1k (target) adaptation, CropAT achieves 52.2% mAP, surpassing the baseline method and the previous SOTA by 6.5% and 3.1%.

Video editing with diffusion models presents significant challenges, especially under real-time constraints. Current methods either enhance temporal consistency at the cost of slow processing or rely on frame-by-frame editing, leading to flickering and temporal artifacts. To address both challenges, we propose SmoothDiffusion-VE, a streaming-based editing approach that improves temporal consistency and processing speed through our proposed Adaptive Feature Cache (AFC) and motion-guided attention. The AFC dynamically adjusts the caching behavior based on perceptual similarity (LPIPS) between frames, i.e., shifting to a mini-cache mode for similar frames to reduce computational load. Conversely, significant frame changes trigger deeper caching to maintain robust temporal coherence. Our motion-guided attention selectively focuses on dynamic regions using optical flow, reducing unnecessary computations in static areas and accelerating processing. SmoothDiffusion-VE can run 28 FPS on one RTX 4090 GPU, achieving a 1564xspeedup over Plug-and-Play Diffusion (PNP) and a 1916xspeedup over Diffusion Motion Transfer (DMT), delivering a powerful solution for fast and consistent video editing.

Robust visual recognition on embedded platforms requires models that both generalize out-of-distribution (OOD) and fit into tiny compute/memory budgets. While pre-training is a standard route to robustness for mid/large backbones, its value in the ultra-small regime remains unclear. We present a capacity-aware study of pre-training for two efficient ConvNet families (EfficientNet and MobileNetV3) scaled from "small" to "ultra-small" via a simple, reproducible recipe. We compare three initializations -- ImageNet->COCO pretraining, ImageNet classification pretraining, and training from scratch--on two axes of distribution shift: (i) cross-dataset RGB-RGB transfer between LLVIP and FLIR (ii) cross-modality detection where models are fine-tuned on RGB and evaluated on infrared (IR). A complementary classification study on DomainNet probes whether the trends extend beyond detection. Across settings, we find that pretraining's benefit is conditional on both backbone capacity and shift difficulty. Task-aligned Imagenet->COCO pretraining is the most reliable starting point at moderate sizes and for the easier transfer direction. In the low-capacity regimes, differences are typically within run-to-run variation, and training from scratch can match or surpass pre-training. Classification mirrors this capacity gating. Our results test the premise "pretraining always helps" and instead quantify when task-aligned pretraining pays off for ultra-small backbones and when it likely does not(The code will be available online after acceptance.).

Video anomaly understanding (VAU) aims to provide detailed interpretation and semantic comprehension of anomalous events within videos, addressing limitations of traditional methods that focus solely on detecting and localizing anomalies. However, existing approaches often neglect the deeper causal relationships and interactions between objects, which are critical for understanding anomalous behaviors. In this paper, we propose VADER, an LLM-driven framework for Video Anomaly unDErstanding, which integrates keyframe object Relation features with visual cues to enhance anomaly comprehension from video. Specifically, VADER first applies an Anomaly Scorer to assign per-frame anomaly scores, followed by a Context-AwarE Sampling (CAES) strategy to capture the causal context of each anomalous event. A Relation Feature Extractor and a COntrastive Relation Encoder (CORE) jointly model dynamic object interactions, producing compact relational representations for downstream reasoning. These visual and relational cues are integrated with LLMs to generate detailed, causally grounded descriptions and support robust anomaly-related question answering. Experiments on multiple real-world VAU benchmarks demonstrate that VADER achieves strong results across anomaly description, explanation, and causal reasoning tasks, advancing the frontier of explainable video anomaly analysis.

System designers and developers need data-driven approaches for user-interface (UI) development and testing. They need trace-based workflows to support UI navigation agents and inputs for UI code generation. Given the high production costs of manually constructing such workflows, the UI agent research community has explored automated, fully-synthetic UI workflow generation. However, there is an open need to characterize the fidelity and effectiveness of these synthetic approaches with respect to the current manual approaches.In this work, we aim to provide larger-scale synthetic workflows based on past human usage. We particularly focused on the desktop application modality since prior synthetic generation has mostly targeted mobile or web applications. Using video tutorials with permissive licenses, we derive associated UI behaviors to construct synthetic dataset of UI workflows in desktop applications. We provide data from these videos to a large language model (LLM) to generate a set of "task list" instructions that replicate the videos' actions. We execute the task list instructions in a UI agent within an instrumented desktop operating system. Using the sensor data from that environment, we can enable simpler actuation scripts for replay of the tasks. Along with detailing our approach, we provide a dataset with over 5,000 workflows across a range of desktop applications with costs that are roughly 51% to 70% of manual construction.

Low signal-to-noise ratio (SNR) videos -- such as those from underwater sonar, ultrasound, and microscopy -- pose significant challenges for computer vision models, particularly when paired clean imagery for denoising is unavailable. We present Spatiotemporal Augmentations and denoising in Video for Downstream Tasks (SAVeD), a novel self-supervised method that denoises low-SNR sensor videos using only raw noisy data. By leveraging distinctions between foreground and background motion and exaggerating objects with stronger motion signal, SAVeD enhances foreground object visibility and reduces background and camera noise without requiring clean video. SAVeD has a set of architectural optimizations that lead to faster throughput, training, and inference than existing deep learning methods. We also introduce a new denoising metric, FBD, which indicates foreground-background divergence for detection datasets without requiring clean imagery. Our approach achieves state-of-the-art results for classification, detection, tracking, and counting tasks and it does so with fewer training resource requirements than existing deep-learning-based denoising methods.

Vision Graph Neural Networks (ViGs) offer a new direction for advancements in vision architectures. While powerful, ViGs often face substantial computational challenges stemming from their graph construction phase, which can hinder their efficiency. To address this issue we propose AdaptViG, an efficient and powerful hybrid Vision GNN that introduces a novel graph construction mechanism called Adaptive Graph Convolution. This mechanism builds upon a highly efficient static axial scaffold and a dynamic, content-aware gating strategy called Exponential Decay Gating. This gating mechanism selectively weighs long-range connections based on feature similarity. Furthermore, AdaptViG employs a hybrid strategy, utilizing our efficient gating mechanism in the early stages and a full Global Attention block in the final stage for maximum feature aggregation. Our method achieves a new state-of-the-art trade-off between accuracy and efficiency among Vision GNNs. For instance, our AdaptViG-M achieves 82.6% top-1 accuracy, outperforming ViG-B by 0.3% while using 80% fewer parameters and 84% fewer GMACs. On downstream tasks, AdaptViG-M obtains 45.8 mIoU, 44.8 APbox, and 41.1 APmask, surpassing the much larger EfficientFormer-L7 by 0.7 mIoU, 2.2 APbox, and 2.1 APmask, respectively, with 78% fewer parameters.

Unmanned Aerial Vehicles (UAVs) are increasingly deployed in close proximity to humans for applications such as parcel delivery, traffic monitoring, disaster response and infrastructure inspections. Ensuring safe and reliable operation in these human-populated environments demands accurate perception of human poses and actions from an aerial viewpoint. This perspective challenges existing methods with low resolution, steep viewing angles and (self-)occlusion, especially if the application demands realtime feasibile models. We train and deploy FlyPose, a lightweight top-down human pose estimation pipeline for aerial imagery. Through multi-dataset training, we achieve an average improvement of 6.8 mAP in person detection across the test-sets of Manipal-UAV, VisDrone, HIT-UAV as well as our custom dataset. For 2D human pose estimation we report an improvement of 16.3 mAP on the challenging UAV-Human dataset. FlyPose runs with an inference latency of 20 milliseconds including preprocessing on a Jetson Orin AGX Developer Kit and is deployed onboard a quadrotor UAV during flight experiments. We also publish FlyPose-104, a small but challenging aerial human pose estimation dataset, that includes manual annotations from difficult aerial perspectives: https://github.com/farooqhassaan/FlyPose.

Modern deep neural networks have shown remarkable performance in medical image classification. However, such networks either emphasize pixel-intensity features instead of fundamental anatomical structures (e.g., those encoded by topological invariants), or they capture only simple topological features via single-parameter persistence.In this paper, we propose a new topology-guided classification framework that extracts multi-scale and multi-filtration persistent topological features and integrates them into vision classification backbones.For an input image, we first compute cubical persistence diagrams (PDs) across multiple image resolutions/scales. We then develop a "vineyard" algorithm that consolidates these PDs into a single, stable diagram capturing signatures at varying granularities, from global anatomy to subtle local irregularities that may indicate early-stage disease.To further exploit richer topological representations produced by multiple filtrations, we design a cross-attention-based neural network that directly processes the consolidated final PDs. The resulting topological embeddings are fused with feature maps from CNNs or Transformers. By integrating multi-scale and multi-filtration topologies into an end-to-end architecture, our approach enhances the model's capacity to recognize complex anatomical structures.Evaluations on three public datasets show consistent, considerable improvements over strong baselines and state-of-the-art methods, demonstrating the value of our comprehensive topological perspective for robust and interpretable medical image classification.

Maps are powerful carriers of structured and contextual knowledge, encompassing geography, demographics, infrastructure, and environmental patterns. Reasoning over such knowledge requires models to integrate spatial relationships, visual cues, real-world context, and domain-specific expertise--capabilities that current large language models (LLMs) and vision-language models (VLMs) still struggle to exhibit consistently. Yet, datasets used to benchmark VLMs on map-based reasoning remain narrow in scope, restricted to specific domains, and heavily reliant on artificially generated content (outputs from LLMs or pipeline-based methods), offering limited depth for evaluating genuine geospatial reasoning.To address this gap, we present MapVerse, a large-scale benchmark built on real-world maps. It comprises 11,837 human-authored question-answer pairs across 1,025 maps, spanning ten diverse map categories and multiple question categories for each. The dataset provides a rich setting for evaluating map reading, interpretation, and multimodal reasoning.We evaluate ten state-of-the-art models against our benchmark to establish baselines and quantify reasoning gaps. Beyond overall performance, we conduct fine-grained categorical analyses to assess model inference across multiple dimensions and investigate the visual factors shaping reasoning outcomes. Our findings reveal that while current VLMs perform competitively on classification-style tasks, both open- and closed-source models fall short on advanced tasks requiring complex spatial reasoning. All codes and data will be publicly released via github upon acceptance.

Detecting visual anomalies in diverse, multi-class real-world images is a significant challenge. We introduce VLMDiff, a novel unsupervised multi-class visual anomaly detection framework. It integrates a Latent Diffusion Model (LDM) with a Vision-Language Model (VLM) for enhanced anomaly localization and detection. Specifically, a pre-trained VLM with a simple prompt extracts detailed image descriptions, serving as additional conditioning for LDM training. Current diffusion-based methods rely on synthetic noise generation, limiting their generalization and requiring per-class model training, which hinders scalability. VLMDiff, however, leverages VLMs to obtain normal captions without manual annotations or additional training. These descriptions condition the diffusion model, learning a robust normal image feature representation for multi-class anomaly detection. Our method achieves competitive performance, improving the pixel-level Per-Region-Overlap (PRO) metric by up to 25 points on the Real-IAD dataset and 8 points on the COCO-AD dataset, outperforming state-of-the-art diffusion-based approaches. Code is available at https://github.com/giddyyupp/VLMDiff.

Neural networks are frequently used in medical diagnosis. However, due to their black-box nature, model explainers are used to help clinicians understand better and trust model outputs. This paper introduces an explainer method for classifying Retinopathy of Prematurity (ROP) from fundus images. Previous methods fail to generate explanations that preserve input image structures such as smoothness and sparsity. We introduce Sparse and Smooth Explainer (SSplain), a method that generates pixel-wise explanations while preserving image structures by enforcing smoothness and sparsity. This results in realistic explanations to enhance the understanding of the given black-box model. To achieve this goal, we define an optimization problem with combinatorial constraints and solve it using the Alternating Direction Method of Multipliers (ADMM). Experimental results show that SSplain outperforms commonly used explainers in terms of both post-hoc accuracy and smoothness analyses. Additionally, SSplain identifies features that are consistent with domain-understandable features that clinicians consider as discriminative factors for ROP. We also show SSplain's generalization by applying it to additional publicly available datasets. Code is available at https://github.com/neu-spiral/SSplain.

Deep learning-based vision systems are increasingly deployed in high-stakes applications, yet remain vulnerable to imperceptible manipulations that exploit model blind spots. We present BAFLE-DCT, a frequency-domain steganography framework that achieves high-capacity, imperceptible data embedding while evading state-of-the-art deep steganalysis. Unlike traditional spatial-domain methods that alter pixel values and trigger visual or statistical artifacts, BAFLE-DCT operates in the DCT domain, selectively modifying mid-frequency coefficients in perceptually insignificant regions identified via saliency analysis. A lightweight feedforward network further refines block selection using entropy and DCT variance features to balance embedding capacity and visual fidelity. Stego images generated by BAFLE-DCT consistently bypass advanced detectors such as YeNet and SRNet, yielding near-random detection rates ( 50%) across payload sizes. Importantly, embedded images maintain classification consistency under CLIP, demonstrating semantic preservation. We also release a large-scale, full-color steganographic dataset for frequency-domain research, addressing limitations of grayscale, spatial-domain benchmarks. Our results expose critical vulnerabilities in visual content authentication pipelines and motivate the development of frequency-aware detection strategies.