WACV 2026 Papers

Deep neural networks often inherit social and demographic biases from annotated data during model training, leading to unfair predictions, especially in the presence of sensitive attributes like race, age, gender etc. Existing methods fall prey to the inherent data imbalance between attribute groups and inadvertently emphasize on sensitive attributes, worsening unfairness and performance. To surmount these challenges, we propose SHaSaM (Submodular Hard Sample Mining), a novel combinatorial approach that models fairness-driven representation learning as a submodular hard-sample mining problem. Our two-stage approach comprises of SHaSaM-MINE, which introduces a submodular subset selection strategy to mine hard positives and negatives -- effectively mitigating data imbalance, and SHaSaM-LEARN, which introduces a family of combinatorial loss functions based on Submodular Conditional Mutual Information to maximize the decision boundary between target classes while minimizing the influence of sensitive attributes. This unified formulation restricts the model from learning features tied to sensitive attributes, significantly enhancing fairness without sacrificing performance. Experiments on CelebA and UTKFace demonstrate that SHaSaM achieves state-of-the-art results, with up to 2.7 points improvement in model fairness (Equalized Odds) and a 3.5% gain in Accuracy, within fewer epochs as compared to existing methods.

Class-level machine unlearning has been proposed to address security and privacy issues of deep neural networks (DNNs). However, existing approaches either exhibit low performance or have excessive computation/storage requirements. This makes them inapplicable in mobile computing scenarios, where computation and memory are severely constrained yet unlearning has to be performed frequently and effectively. This limitation is mainly due to the usage of a retain dataset, i.e., a sub-dataset containing the knowledge that the DNN should maintain after the unlearning. In this paper, we propose CLUE, an unlearning algorithm that does not require a retain dataset. Our key idea is to treat inputs coming from the forget class as out-of-distribution data and to use knowledge distillation to impose this constraint on the updated DNN. We have experimentally evaluated CLUE on Resnet-20, ViT-Base, and ViT-Large DNNs trained on CIFAR10, CIFAR100, andVGGFace2 datasets. We have also implemented CLUE on Raspberry PI and compared the power consumption and latency of CLUE with respect to several existing baselines. We show that CLUE improves power consumption by 68% and latency by 90% while improving the unlearning performance by up to 4.74%.

This study examines the feasibility of employing raw two-dimensional gas chromatography/time-of-flight mass spectrometry (GCxGC ToF-MS) data for the purpose of human scent identity verification. Unlike techniques that require expert-driven identification of compounds, our framework transforms each GCxGC sample into a multi-channel image. A comprehensive assessment has been conducted on ten channel-encoding schemes, five spatial-alignment strategies, and ten feature-embedding methods.The evaluation is performed on a newly assembled dataset of 252 individuals, comprising 2,528 raw samples and aggregating around 7.5TB of data. In contrast to conventional methodologies employed in chemical analysis, our research demonstrates that alignment to a common spatial reference frame is unnecessary. The best performing method reaches an approximately 53% true positive rate at a 5% false positive rate. Although this performance is below that of well-established biometrics (e.g., iris verification), our results underscore the feasibility of raw-odor-based verification for scenarios where direct line-of-sight or cooperation may be limited, thereby revealing opportunities for interdisciplinary research.We will release the code and datasets with the camera ready.

Video Large Multimodal Models (VLMMs) have shown impressive performance in video understanding, yet their ability to accurately capture the temporal order of multiple events remains underexplored. We interestingly observe that, even when video frames are scrambled, models perform very well on the existing benchmarks by comprehensive experiments. This implies that VLMMs may not necessarily rely on accurate sequential processing of visual events, but instead depend on prior knowledge of typical scenarios to answer the question. To benchmark temporal understanding capabilities in VLMMs, we propose VECTOR, designed to explicitly assess a model's ability to identify the temporal order of events. On this benchmark, we observe that various VLMMs often fail to understand the orders of events. To address this, we propose MECOT (Multi-Event instruction fine-tuning with Chain-of-Thought), which (1) trains models on detailed, event-by-event video descriptions and (2) using chain-of-thought prompts at inference to enhance temporal awareness. MECOT outperforms prior arts on VECTOR as well as improving performance on existing video benchmarks, implying effectiveness of temporal understanding.

This paper presents QuEENet, a hybrid quantum-classical architecture for image classification that incorporates parameterized quantum circuits within a convolution neural network. This study investigates how quantum circuit expressivity and entanglement strategies influence classification performance, with a focus on configurations involving a CNOT gate followed by a rotational gate on the target qubit. Non-Clifford gates, such as Rx/Ry/Rz supports larger state-space coverage and expressivity in quantum models. The proposed QuEENet explored the aspect of non-Clifford gates in parameterized quantum circuits. While non-Clifford gates are theoretically critical for universal quantum computation, but their role in image classification task is unexplored. Experimental results across multiple benchmark datasets suggest that while increased expressivity via non-Clifford gates can be beneficial, it should be carefully balanced with circuit interpretability and trainability. QuEENet demonstrates that hybrid models can leverage quantum circuits not merely as architectural novelties, but as controllable modules for enhancing learning in classical pipelines. An extensive ablation study was conducted across multiple datasets to highlight the effects of Clifford and non-Clifford gate combinations and entanglement configurations.

Digital Terrain Models (DTMs) represent the bare-earth elevation and are important in numerous geospatial applications. Such data models cannot be directly measured by sensors and are typically generated from Digital Surface Models (DSMs) derived from LiDAR or photogrammetry. Traditional filtering approaches rely on manually tuned parameters, while learning-based methods require well-designed architectures, often combined with post-processing. To address these challenges, we introduce Ground Diffusion (GrounDiff), the first diffusion-based framework that iteratively removes non-ground structures by formulating the problem as a denoising task. We incorporate a gated design with confidence-guided generation that enables selective filtering. To increase scalability, we further propose Prior-Guided Stitching (PrioStitch), which employs a downsampled global prior automatically generated using GrounDiff to guide local high-resolution predictions. We evaluate our method on the DSM-to-DTM translation task across diverse datasets, showing that GrounDiff consistently outperforms deep learning-based state-of-the-art methods, reducing RMSE by up to 93% on ALS2DTM and up to 47% on USGS benchmarks. In the task of road reconstruction, which requires both high precision and smoothness, our method achieves up to 81% lower distance error compared to specialized techniques on the GeRoD benchmark, while maintaining competitive surface smoothness using only DSM inputs, without task-specific optimization. Our variant for road reconstruction, GrounDiff+, is specifically designed to produce even smoother surfaces, further surpassing state-of-the-art methods. The project page is available at https://deepscenario.github.io/GrounDiff/ .

Recent object detectors have achieved impressive accuracy in identifying objects seen during training. However, real-world deployment often introduces novel and unexpected objects, referred to as out-of-distribution (OOD) objects, posing significant challenges to model trustworthiness. Modern object detectors are typically overconfident, making it unreliable to use their predictions *alone* for OOD detection. To address this, we propose leveraging an *auxiliary* model as a complementary solution. Specifically, we utilize an off-the-shelf text-to-image generative model, such as Stable Diffusion, which is trained with objective functions distinct from those of discriminative object detectors. We hypothesize that this fundamental difference enables the detection of OOD objects by measuring inconsistencies between the models. Concretely, for a given detected object bounding box and its predicted in-distribution class label, we perform class-conditioned inpainting on the image with the object removed. If the object is OOD, the inpainted image is likely to deviate significantly from the original, making the reconstruction error a robust indicator of OOD status. Extensive experiments demonstrate that our approach consistently surpasses existing zero-shot and non-zero-shot OOD detection methods, establishing a robust framework for enhancing object detection systems in dynamic environments.

Medical image segmentation (MIS) is challenged by anatomical variability, ambiguous boundaries, and subtle textures, demanding an efficient balance between fine local details and global context. Existing architectures often suffer from suboptimal fusion of spatial and frequency-domain features, limiting their ability to capture richer structural and textural representations. To specifically address these challenges, we introduce FLoMo-Net, a modular MIS architecture organized around a principled route -> select -> refine -> correct pipeline: (1) The Local-Global Mixture of Experts encoder adaptively routes features across specialized convolutional branches to capture scale-appropriate context; (2) The Dual-Attention Selective Aggregator then jointly selects informative channels and spatial regions using frequency-guided modulation; (3) At the bridge, the Frequency-Aware Multi-Scale Refinement module refines edges and textures through explicit low/high-frequency decomposition; (4) Finally, the False Positive/Negative Corrective Attention Module leverages uncertainty, derived from entropy and cosine dissimilarity, to produce a residual corrective mask that suppresses semantic drift and improves boundary delineation in the decoder stages. Across four MIS benchmarks, FLoMo-Net achieves superior boundary-aware performance and faster inference with fewer parameters than prior state-of-the-art 2D MIS models. Code is publicly available at https://github.com/rayhan-ahmed91/FLoMo-Net.

Deep generative models have emerged as a transformative tool in medical imaging, offering substantial potential for synthetic data generation. However, recent empirical studies highlight a critical vulnerability: these models can memorize sensitive training data, posing significant risks of unauthorized patient information disclosure. Detecting memorization in generative models remains particularly challenging, necessitating scalable methods capable of identifying training data leakage across large sets of generated samples. In this work, we propose DeepSSIM, a novel self-supervised metric for quantifying memorization in generative models. DeepSSIM is trained to: i) project images into a learned embedding space and ii) force the cosine similarity between embeddings to match the ground-truth Structural Similarity Index (SSIM) scores computed in the image space. To capture domain-specific anatomical features, training incorporates structure-preserving augmentations, allowing DeepSSIM to estimate similarity reliably without requiring precise spatial alignment. We evaluate DeepSSIM in two case studies using synthetic brain MRI and chest X-ray data generated by a Latent Diffusion Model (LDM) trained under memorization-prone conditions. Compared to state-of-the-art memorization metrics, DeepSSIM achieves superior performance, improving F1 scores by an average of +52.03% over the best existing method. Code and data are publicly available at https://github.com/brAIn-science/DeepSSIM.

Recent advances in Source-Free Unsupervised Video Domain Adaptation (SFUVDA) leverage vision-language models to enhance pseudo-label generation. However, challenges such as noisy pseudo-labels and over-confident predictions limit their effectiveness in adapting well across domains. We propose Co-STAR, a novel framework that integrates curriculum learning with collaborative self-training between a source-trained teacher and a contrastive vision-language model (CLIP). Our curriculum learning approach employs a reliability-based weight function that measures bidirectional prediction alignment between the teacher and CLIP, balancing between confident and uncertain predictions. This function preserves uncertainty for difficult samples, while prioritizing reliable pseudo-labels when the predictions from both models closely align. To further improve adaptation, we propose Adaptive Curriculum Regularization, which modifies the learning priority of samples in a probabilistic, adaptive manner based on their confidence scores and prediction stability, mitigating overfitting to noisy and over-confident samples. Extensive experiments across multiple video domain adaptation benchmarks demonstrate that Co-STAR consistently outperforms state-of-the-art SFUVDA methods. Code is available at https://github.com/Plrbear/Co-Star

Whole slide images, with their gigapixel-scale panoramas of tissue samples, are pivotal for precise disease diagnosis. However, their analysis is hindered by immense data size and scarce annotations. Existing MIL methods face challenges due to the fundamental imbalance where a single bag-level label must guide the learning of numerous patch-level features. This sparse supervision makes it difficult to reliably identify discriminative patches during training, leading to unstable optimization and suboptimal solutions. We propose a spatially regularized MIL framework that leverages inherent spatial relationships among patch features as label-independent regularization signals. Our approach learns a shared representation space by jointly optimizing feature-induced spatial reconstruction and label-guided classification objectives, enforcing consistency between intrinsic structural patterns and supervisory signals. Experimental results on multiple public datasets demonstrate significant improvements over state-of-the-art methods, offering a promising direction for pathology image analysis.

Pose-Guided Person Image Synthesis (PGPIS) aims to transfer a person from a source image to a target pose (e.g., skeleton) while preserving their original appearance. Although existing methods can produce high-quality results at first glance, they often suffer from noticeable distortions in fine details. We identify the root cause of these issues as the heavy reliance on pre-trained encoders for extracting visual features from the source image. To address this, we propose a novel Query Enhancement Network composed of two key components: the Query-based Feature Fusion Transformer (QFFT) and Pose-Masked Attention (PMA). The QFFT uses learnable queries to fuses multi-scale features from high to low resolution extracted by the backbone encoder, thereby significantly enhancing the realism of texture details in the generated images. To better capture the relationship between pose information and visual features from the source image, we introduce PMA that uses the pose skeleton as a mask to guide the attention mechanism to focus on the pose regions. Our method produces high-quality, visually coherent results and outperforms existing approaches on standard evaluation metrics, including FID, SSIM, and LPIPS, demonstrating its effectiveness on the DeepFashion dataset.

Reliable operation of wind turbines requires frequent inspections, as even minor surface damages can degrade aerodynamic performance, reduce energy output, and accelerate blade wear. Central to automating these inspections is the accurate segmentation of turbine blades from visual data. This task is traditionally addressed through dense, pixel-wise deep learning models. However, such methods demand extensive annotated datasets, posing scalability challenges. In this work, we introduce an annotation-efficient segmentation approach that reframes the pixel-level task into a binary region classification problem. Image regions are generated using a fully unsupervised, interpretable Modular Adaptive Region Growing technique, guided by image-specific Adaptive Thresholding and enhanced by a Region Merging process that consolidates fragmented areas into coherent segments. To improve generalization and classification robustness, we introduce RegionMix, an augmentation strategy that synthesizes new training samples by combining distinct regions. Our framework demonstrates state-of-the-art segmentation accuracy and strong cross-site generalization by consistently segmenting turbine blades across distinct windfarms.

Marine visual understanding is essential for monitoring and protecting marine ecosystems, enabling automatic and scalable biological surveys. However, progress is hindered by limited training data and the lack of a systematic task formulation that aligns domain-specific marine challenges with well-defined computer vision tasks, thereby limiting effective model application. To address this gap, we present Orca, a multi-modal benchmark for marine research comprising 14,647 images from 478 species, with 42,217 bounding box annotations and 22,321 expert-verified instance captions. The dataset provides fine-grained visual and textual annotations that capture morphology-oriented attributes across diverse marine species. To catalyze methodological advances, we evaluate 18 state-of-the-art models on three tasks: object detection (closed-set and open-vocabulary), instance captioning, and visual grounding. Results highlight key challenges, including species diversity, morphological overlap, and specialized domain demands, underscoring the difficulty of marine understanding. Orca thus establishes a comprehensive benchmark to advance research in marine domain.

Wilderness Search and Rescue (WiSAR) operations aim to locate and rescue individuals in remote, rugged environments. We introduce WiSAR3D, the first aerial LiDAR 3D dataset specifically tailored for 3D object detection in rural scenes. WiSAR3D comprises 2633 strips---contiguous point clouds collected along the flight path, all fully annotated with 67.5K 3D bounding boxes for 22 categories. As the data is captured with multiple returns (echoes), it enables detection under foliage, which is impossible with RGB cameras. Additionally, as this is the first-of-its-kind dataset, we provide benchmark evaluations for state-of-the-art methods developed for autonomous cars on our aerial dataset. The results suggest that specialized models are needed in this domain, as finding individuals remains challenging for current 3D detectors. The code and dataset are publicly available at https://github.com/oshrout/WiSAR3D.

Scene text spotting aims to detect and recognize text in real-world images, where instances are often short, fragmented, or visually ambiguous. Existing methods primarily rely on visual cues and implicitly capture local character dependencies, but they overlook the benefits of external linguistic knowledge. Prior attempts to integrate language models either adapt language modeling objectives without external knowledge or apply pretrained models that are misaligned with the word-level granularity of scene text. We propose TiCLS, an end-to-end text spotter that explicitly incorporates external linguistic knowledge from a character-level pretrained language model. TiCLS contains a pretrained linguistic decoder that fuses visual and linguistic features, enabling robust recognition of ambiguous or fragmented text. Experiments on ICDAR 2015, Total-Text, and CTW1500 demonstrate that TiCLS achieves state-of-the-art performance, validating the effectiveness of PLM-guided linguistic integration for scene text spotting. The code is available at https://github.com/knowledge-computing/TiCLS

The quality of captured images significantly impacts the performance of downstream perception tasks. Recent works that co-design camera systems with perception tasks have demonstrated improved task performance. However, prior approaches primarily focus on optimising fixed camera parameters determined at manufacturing, whereas many parameters, such as exposure settings, require adaptive control at runtime. This paper presents a method that jointly optimises camera hardware and adaptive camera control algorithms alongside downstream vision tasks. We propose a unified optimisation framework that combines gradient-based and derivative-free methods to support continuous and discrete parameters, non-differentiable image formation processes, and a neural network-based adaptive camera control algorithm. To handle non-differentiable rendering of some image effects, such as motion blur, we propose DF-Grad, a hybrid optimisation method that supervises the neural adaptive control algorithm using signals from the derivative-free optimiser, in addition to unsupervised task-driven learning. Experiments show that the proposed method outperforms baselines that optimise camera hardware and camera control algorithm separately, particularly under challenging conditions such as low light and fast motion. We demonstrate that joint optimisation of both static camera hardware parameters and adaptive control algorithms leads to improved perception task performance, offering a unified approach to task-driven camera system design.

Grounding complex, compositional visual queries with multiple objects and relationships is a fundamental challenge for vision-language models. While standard phrase grounding methods excel at localizing single objects, they lack the structural inductive bias to parse intricate relational descriptions, often failing as queries become more descriptive. To address this structural deficit, we focus on scene-graph grounding, a powerful but less-explored formulation where the query is an explicit graph of objects and their relationships. However, existing methods for this task also struggle, paradoxically showing decreased performance as the query graph grows---failing to leverage the very information that should make grounding easier. We introduce SceneProp, a novel method that resolves this issue by reformulating scene-graph grounding as a Maximum a Posteriori (MAP) inference problem in a Markov Random Field (MRF). By performing global inference over the entire query graph, SceneProp finds the optimal assignment of image regions to nodes that jointly satisfies all constraints. This is achieved within an end-to-end framework via a differentiable implementation of the Belief Propagation algorithm. Experiments on four benchmarks show that our dedicated focus on the scene-graph grounding formulation allows SceneProp to significantly outperform prior work. Critically, its accuracy consistently improves with the size and complexity of the query graph, demonstrating for the first time that more relational context can, and should, lead to better grounding. Codes are available at https://github.com/keitaotani/SceneProp.

Generating 3D human motions from textual descriptions is an important research problem with broad applications in video games, virtual reality, and augmented reality. Recent methods align the textual description with human motion at the sequence level, neglecting the internal semantic structure of modalities. However, both motion descriptions and motion sequences can be naturally decomposed into smaller and semantically coherent segments, which can serve as atomic alignment units to achieve finer-grained correspondence. Motivated by this, we propose SegMo, a novel Segment-aligned text-conditioned human Motion generation framework to achieve fine-grained text-motion alignment. Our framework consists of three modules: (1) Text Segment Extraction, which decomposes complex textual descriptions into temporally ordered phrases, each representing a simple atomic action; (2) Motion Segment Extraction, which partitions complete motion sequences into corresponding motion segments; and (3) Fine-grained Text-Motion Alignment, which aligns text and motion segments with contrastive learning. Extensive experiments demonstrate that SegMo improves the strong baseline on two widely used datasets, achieving an improved TOP 1 score of 0.553 on the HumanML3D test set. Moreover, thanks to the learned shared embedding space for text and motion segments, SegMo can also be applied to retrieval-style tasks such as motion grounding and motion-to-text retrieval.