CVPR 2025 Papers

Image steganography can hide information in a host image and obtain a stego image that is perceptually indistinguishable from the original one. This technique has tremendous potential in scenarios like copyright protection and information retrospection. Some previous studies have proposed to enhance the robustness of the methods against image disturbances to increase their applicability. However, they generally cannot achieve a satisfying balance between the steganography quality and robustness. Instead of image-in-image steganography, we focus on the issue of message-in-image embedding that is robust to various real-world image distortions. This task aims to embed information into a natural image and the decoding result is required to be completely accurate, which increases the difficulty of data concealing and revealing. Inspired by the recent developments in transformer-based vision models, we discover that the tokenized representation of image is naturally suitable for steganography task. In this paper, we propose a novel message embedding framework, called Robust Message Steganography (RMSteg), which is competent to hide message via QR Code in a host image based on an normalizing flow-based model. The stego image derived by our method has imperceptible changes and the encoded message can be accurately restored even if the image is printed out and photographed. To our best knowledge, this is the first work that integrates the advantages of transformer models into normalizing flow. The code is available at https://github.com/huayuan4396/RMSteg.

Recently, natural language has been the primary medium for human-robot interaction. However, its inherent lack of spatial precision for robotic control introduces challenges such as ambiguity and verbosity. To address these limitations, we introduce the ***Robotic Visual Instruction (RoVI)***, a novel paradigm to guide robotic tasks through an object-centric, hand-drawn symbolic representation. RoVI effectively encodes spatial-temporal information into human-interpretable visual instructions through 2D sketches, utilizing arrows, circles, colors, and numbers to direct 3D robotic manipulation. To enable robots to understand RoVI better and generate precise actions based on RoVI, we present ***Visual Instruction Embodied Workflow (VIEW)***, a pipeline formulated for RoVI-conditioned policies. This approach leverages Vision-Language Models (VLMs) to interpret RoVI inputs, decode spatial and temporal constraints from 2D pixel space via keypoint extraction, and then transform them into executable 3D action sequences. We additionally curate a specialized dataset of 15K instances to fine-tune small VLMs for edge deployment, enabling them to effectively learn RoVI capabilities. Our approach is rigorously validated across 11 novel tasks in both real and simulated environments, demonstrating significant generalization capability. Notably, VIEW achieves an 87.5% success rate in real-world scenarios involving unseen tasks that feature multi-step actions, with disturbances, and trajectory-following requirements. Code and Datasets in this paper will be released soon.

Spatial understanding is a crucial capability that enables robots to perceive their surroundings, reason about their environment, and interact with it meaningfully. In modern robotics, these capabilities are increasingly provided by vision-language models. However, these models face significant challenges in spatial reasoning tasks, as their training data are based on general-purpose image datasets that often lack sophisticated spatial understanding. For example, datasets frequently do not capture reference frame comprehension, yet effective spatial reasoning requires understanding whether to reason from ego-, world-, or object-centric perspectives. To address this issue, we introduce RoboSpatial, a large-scale dataset for spatial understanding in robotics. It consists of real indoor and tabletop scenes, captured as 3D scans and egocentric images, and annotated with rich spatial information relevant to robotics. The dataset includes 1M images, 5k 3D scans, and 3M annotated spatial relationships, and the pairing of 2D egocentric images with 3D scans makes it both 2D- and 3D- ready. Our experiments show that models trained with RoboSpatial outperform baselines on downstream tasks such as spatial affordance prediction, spatial relationship prediction, and robotics manipulation.

Vision-based pose estimation of articulated robots with unknown joint angles has applications in collaborative robotics and human-robot interaction tasks. Current frameworks use neural network encoders to extract image features and downstream layers to predict joint angles and robot pose. While images of robots inherently contain rich information about the robot's physical structures, existing methods often fail to leverage it fully; therefore, limiting performance under occlusions and truncations. To address this, we introduce RoboPEPP, a method that fuses information about the robot's physical model into the encoder using a masking-based self-supervised embedding-predictive architecture. Specifically, we mask the robot's joints and pre-train an encoder-predictor model to infer the joints' embeddings from surrounding unmasked regions, enhancing the encoder's understanding of the robot's physical model. The pre-trained encoder-predictor pair, along with joint angle and keypoint prediction networks, is then fine-tuned for pose and joint angle estimation. Random masking of input during fine-tuning and keypoint filtering during evaluation further improves robustness. Our method, evaluated on several datasets, achieves the best results in robot pose and joint angle estimation while being the least sensitive to occlusions and requiring the lowest execution time.

Recent advancements in robotic manipulation have highlighted the potential of intermediate representations for improving policy generalization. In this work, we explore grounding masks as an effective intermediate representation, balancing two key advantages: (1) effective spatial guidance that specifies target objects and placement areas while also conveying information about object shape and size, and (2) broad generalization potential driven by large-scale vision-language models pretrained on diverse grounding datasets. We introduce \method, a grounding-aware robotic manipulation policy that leverages grounding masks as an intermediate representation to guide policy networks in object manipulation tasks. To further explore and enhance generalization, we propose an automated pipeline for generating large-scale, simulated data with a diverse set of objects and instructions. Extensive experiments show the value of our dataset and the effectiveness of grounding masks as intermediate guidance, significantly enhancing the generalization abilities of robot policies.

Foundation models for remote sensing have garnered increasing attention for their strong performance across various observation tasks. However, current models lack robustness in managing diverse input types and handling incomplete data in downstream tasks. In this paper, we propose RobSense, a robust multi-modal foundation model for Multi-spectral and Synthetic Aperture Radar data. RobSense is designed with modular components and pre-trained by a combination of temporal multi-modal alignment and masked autoencoder strategies on a huge-scale dataset. Therefore, it can effectively support diverse input types, from static to temporal, uni-modal to multi-modal. To further handle the incomplete data, we incorporate two uni-modal latent reconstructors that recover rich representations from incomplete inputs, addressing variability in spectral bands and temporal sequence irregularities. Extensive experiments demonstrate that RobSense consistently outperforms state-of-the-art baselines on complete datasets across four input types for segmentation, classification, and change detection. On incomplete datasets, RobSense outperforms the baselines by considerably larger margins when the missing rate increases. Project page: https://ikhado.github.io/robsense/

Quality degradation from video compression manifests both spatially along texture edges and temporally with continuous motion changes. Despite recent advances, extracting aligned spatiotemporal information from adjacent frames remains challenging. This is mainly due to limitations in receptive field size and computational complexity, which makes existing methods struggle to efficiently enhance video quality. To address this issue, we propose RivuletMLP, an MLP-based network architecture. Specifically, our framework first employs a Dynamically Guided Deformable Alignment (DDA) module to adaptively explore and align multi-frame feature information. Subsequently, we introduce two modules for feature reconstruction: a Spatiotemporal Feature Flow (SFF) Module and a Benign Selection Compensation (BSC) module. The SFF module establishes non-local dependencies through an innovative feature permutation mechanism. Additionally, the BSC module utilizes a collaborative strategy of deep feature extraction and local region refinement to alleviate inter frame motion discontinuity caused by compression. Experimental results demonstrate that RivuletMLP achieves superior computational efficiency while maintaining powerful reconstruction capabilities.

Self-Supervised Learning (SSL) presents an exciting opportunity to unlock the potential of vast, untapped clinical datasets, for various downstream applications that suffer from the scarcity of labeled data. While SSL has revolutionized fields like natural language processing and computer vision, its adoption in 3D medical image computing has been limited by three key pitfalls: Small pre-training dataset sizes, architectures inadequate for 3D medical image analysis, and insufficient evaluation practices. In this paper, we address these issues by i) leveraging a large-scale dataset of 39k 3D brain MRI volumes and ii) using a Residual Encoder U-Net architecture within the state-of-the-art nnU-Net framework. iii) A robust development framework, incorporating 5 development and 8 testing brain MRI segmentation datasets, allowed performance-driven design decisions to optimize the simple concept of Masked Auto En- coders (MAEs) for 3D CNNs. The resulting model not only surpasses previous SSL methods but also outperforms the strong nnU-Net baseline by an average of approximately 3 Dice points setting a new state-of-the-art. Our code and models are made available here.

Instruction tuning enhances large vision-language models (LVLMs) but increases their vulnerability to backdoor attacks due to their open design. Unlike prior studies in static settings, this paper explores backdoor attacks in LVLM instruction tuning across mismatched training and testing domains. We introduce a new evaluation dimension, backdoor domain generalization, to assess attack robustness under visual and text domain shifts. Our findings reveal two insights: (1) backdoor generalizability improves when distinctive trigger patterns are independent of specific data domains or model architectures, and (2) the competitive interaction between trigger patterns and clean semantic regions, where guiding the model to predict triggers enhances attack generalizability. Based on these insights, we propose a multimodal attribution backdoor attack (MABA) that injects domain-agnostic triggers into critical areas using attributional interpretation. Experiments with OpenFlamingo, Blip-2, and Otter show that MABA significantly boosts the attack success rate of generalization by 36.4% over the unimodal attack, achieving a 97% success rate at a 0.2% poisoning rate. This study reveals limitations in current evaluations and highlights how enhanced backdoor generalizability poses a security threat to LVLMs, even without test data access.

Image restoration aims to recover high-quality (HQ) images from degraded low-quality (LQ) ones by reversing the effects of degradation. Existing generative models for image restoration, including diffusion and score-based models, often treat the degradation process as a stochastic transformation, which introduces inefficiency and complexity. In this work, we propose ResFlow, a novel image restoration framework that models the degradation process as a deterministic path using continuous normalizing flows. ResFlow augments the degradation process with an auxiliary process that disambiguates the uncertainty in HQ prediction to enable reversible modeling of the degradation process. ResFlow adopts entropy-preserving flow paths and learns the augmented degradation flow by matching the velocity field. ResFlow significantly improves the performance and speed of image restoration, completing the task in fewer than four sampling steps. Extensive experiments demonstrate that ResFlow achieves state-of-the-art results across various image restoration benchmarks, offering a practical and efficient solution for real-world applications.

Recent deep-learning-based approaches to single-image reflection removal have shown promising advances, primarily for two reasons: 1) the utilization of recognition-pretrained features as inputs, and 2) the design of dual-stream interaction networks. However, according to the Information Bottleneck principle, high-level semantic clues tend to be compressed or discarded during layer-by-layer propagation. Additionally, interactions in dual-stream networks follow a fixed pattern across different layers, limiting overall performance. To address these limitations, we propose a novel architecture called Reversible Decoupling Network (RDNet), which employs a reversible encoder to secure valuable information while flexibly decoupling transmission- and reflection-relevant features during the forward pass. Furthermore, we customize a transmission-rate-aware prompt generator to dynamically calibrate features, further boosting performance. Extensive experiments demonstrate the superiority of RDNet over existing SOTA methods on five widely-adopted benchmark datasets. Our code will be made publicly available.

Deepfake detection is a long-established research topic vital for mitigating the spread of malicious misinformation. Unlike prior methods that provide either binary classification results or textual explanations separately, we introduce a novel method capable of generating both simultaneously. Our method harnesses the multi-modal learning capability of the pre-trained CLIP and the unprecedented interpretability of large language models (LLMs) to enhance both the generalization and explainability of deepfake detection. Specifically, we introduce a multi-modal face forgery detector (M2F2-Det) that employs tailored face forgery prompt learning, incorporating the pre-trained CLIP to improve generalization to unseen forgeries. Also, M2F2-Det incorporates an LLM to provide detailed textual explanations of its detection decisions, enhancing interpretability by bridging the gap between natural language and subtle cues of facial forgeries. Empirically, we evaluate M2F2-Det on both detection and explanation generation tasks, where it achieves state-of-the-art performance, demonstrating its effectiveness in identifying and explaining diverse forgeries. Source code and models are available at this link.

Class-incremental/Continual image segmentation (CIS) aims to train an image segmenter in stages, where the set of available categories differs at each stage. To leverage the built-in objectness of query-based transformers, which mitigates catastrophic forgetting of mask proposals, current methods often decouple mask generation from the continual learning process. This study, however, identifies two key issues with decoupled frameworks: loss of plasticity and heavy reliance on input data order. To address these, we conduct an in-depth investigation of the built-in objectness and find that highly aggregated image features provide a shortcut for queries to generate masks through simple feature alignment. Based on this, we propose SimCIS, a simple yet powerful baseline for CIS. Its core idea is to directly select image features for query assignment, ensuring "perfect alignment" to preserve objectness, while simultaneously allowing queries to select new classes to promote plasticity. To further combat catastrophic forgetting of categories, we introduce cross-stage consistency in selection and an innovative "visual query"-based replay mechanism. Experiments demonstrate that SimCIS consistently outperforms state-of-the-art methods across various segmentation tasks, settings, splits, and input data orders. All models and codes will be made publicly available at https://github.com/SooLab/SimCIS.

An old-school recipe for training a classifier is to (i) learn a good feature extractor and (ii) optimize a linear layer atop. When only a handful of samples are available per category, as in Few-Shot Adaptation (FSA), data are insufficient to fit a large number of parameters, rendering the above impractical. This is especially true with large pre-trained Vision-Language Models (VLMs), which motivated successful research at the intersection of Parameter-Efficient Fine-tuning (PEFT) and FSA. In this work, we start by analyzing the learning dynamics of PEFT techniques when trained on few-shot data from only a subset of categories, referred to as the "base" classes. We show that such dynamics naturally splits into two distinct phases: (i) task-level feature extraction and (ii) specialization to the available concepts. To accommodate this dynamic, we then depart from prompt- or adapter-based methods and tackle FSA differently. Specifically, given a fixed computational budget, we split it to (i) learn a task-specific feature extractor via PEFT and (ii) train a linear classifier on top. We call this scheme Two-Stage Few-Shot Adaptation (2SFS). Differently from established methods, our scheme enables a novel form of selective inference at a category level, i.e., at test time, only novel categories are embedded by the adapted text encoder, while embeddings of base categories are available within the classifier. Results with fixed hyperparameters across two settings, three backbones, and eleven datasets, show that 2SFS matches or surpasses the state-of-the-art, while established methods degrade significantly across settings.

While vision-language models like CLIP have shown remarkable success in open-vocabulary tasks, their application is currently confined to image-level tasks, and they still struggle with dense predictions. Recent works often attribute such deficiency in dense predictions to the self-attention layers in the final block, and have achieved commendable results by modifying the original query-key attention to self-correlation attention, (e.g., query-query and key-key attention). However, these methods overlook the cross-correlation attention (query-key) properties, which capture the rich spatial correspondence. In this paper, we reveal that the cross-correlation of self-attention in non-final layers of CLIP also exhibits localization properties. Therefore, we propose the Residual Cross-correlation Self-attention (RCS) module, which leverages the cross-correlation self-attention from intermediate layers to remold the attention in the final block. The RCS module effectively reorganizes spatial information, unleashing the localization potential within CLIP for dense vision-language inference. Furthermore, to enhance the focus on regions of the same categories and local consistency, we propose the Semantic Feedback Refinement (SFR) module, which utilizes semantic segmentation maps to further adjust the attention scores. By integrating these two strategies, our method, termed ResCLIP, can be easily incorporated into existing approaches as a plug-and-play module, significantly boosting their performance in dense vision-language inference. Extensive experiments across multiple standard benchmarks demonstrate that our method surpasses state-of-the-art training-free methods, validating the effectiveness of the proposed approach. Code is available at https://github.com/yvhangyang/ResCLIP.

Contrastive Language-Image Pre-training (CLIP) models as a milestone of modern multimodal intelligence, its generalization mechanism grasped massive research interests in the community. While existing studies limited in the scope of pre-training knowledge, hardly underpinned its generalization to countless open-world concepts absent from the pre-training regime. This paper dives into such Out-of-Pre-training (OOP) generalization problem from a holistic perspective. We propose LAION-Beyond benchmark to isolate the evaluation of OOP concepts from pre-training knowledge, with regards to OpenCLIP and its reproducible variants derived from LAION datasets. Empirical analysis evidences that despite image features of OOP concepts born with significant category margins, their zero-shot transfer significantly fails due to the poor image-text alignment. To this, we elaborate the "name-tuning" methodology with its theoretical merits in terms of OOP generalization, then propose few-shot name learning (FSNL) and zero-shot name learning (ZSNL) algorithms to achieve OOP generalization in a data-efficient manner. Their superiority have been further verified in our comprehensive experiments.

Visual localization aims to determine the camera pose of a query image relative to a database of posed images. In recent years, deep neural networks that directly regress camera poses have gained popularity due to their fast inference capabilities. However, existing methods struggle to either generalize well to new scenes or provide accurate camera pose estimates. To address these issues, we present Reloc3r, a simple yet effective visual localization framework. It consists of an elegantly designed relative pose regression network, and a minimalist motion averaging module for absolute pose estimation. Trained on approximately eight million posed image pairs, Reloc3r achieves surprisingly good performance and generalization ability. We conduct extensive experiments on six public datasets, consistently demonstrating the effectiveness and efficiency of the proposed method. It provides high-quality camera pose estimates in real time and generalizes to novel scenes. Code: https://github.com/ffrivera0/reloc3r.

Monocular depth estimation (MDE) models have undergone significant advancements over recent years. Many MDE models aim to predict affine-invariant relative depth from monocular images, while recent developments in large-scale training and vision foundation models enable reasonable estimation of metric (absolute) depth. However, effectively leveraging these predictions for geometric vision tasks, in particular relative pose estimation, remains relatively under explored. While depths provide rich constraints for cross-view image alignment, the intrinsic noise and ambiguity from the monocular depth priors present practical challenges to improving upon classic keypoint-based solutions. In this paper, we develop three solvers for relative pose estimation that explicitly account for independent affine (scale and shift) ambiguities, covering both calibrated and uncalibrated conditions. We further propose a hybrid estimation pipeline that combines our proposed solvers with classic point-based solvers and epipolar constraints. We find that the affine correction modeling is beneficial to not only the relative depth priors but also, surprisingly, the "metric" ones. Results across multiple datasets demonstrate large improvements of our approach over classic keypoint-based baselines and PnP-based solutions, under both calibrated and uncalibrated setups. We also show that our method improves consistently with different feature matchers and MDE models, and can further benefit from very recent advances on both modules.

Neural radiance fields are an emerging 3D scene representation and recently even been extended to learn features for scene understanding by distilling open-vocabulary features from vision-language models. However, current method primarily focus on object-centric representations, supporting object segmentation or detection, while understanding semantic relationships between objects remains largely unexplored. To address this gap, we propose RelationField, the first method to extract inter-object relationships directly from neural radiance fields. RelationField represents relationships between objects as pairs of rays within a neural radiance field, effectively extending its formulation to include implicit relationship queries. To teach RelationField complex, open-vocabulary relationships, relationship knowledge is distilled from multi-modal LLMs. To evaluate RelationField, we solve open-vocabulary 3D scene graph generation tasks and relationship-guided instance segmentation, achieving state-of-the-art performance in both tasks. See the project website at relationfield.github.io.