CVPR 2025 Papers

In this paper, we address the challenging problem of open-world instance segmentation. Existing works have shown that vanilla visual networks are biased toward learning appearance information, e.g. texture, to recognize objects. This implicit bias causes the model to fail in detecting novel objects with unseen textures in the open-world setting. To address this challenge, we propose a learning framework, called view-Consistent LeaRning (v-CLR), which aims to enforce the model to learn appearance-invariant representations for robust instance segmentation. In v-CLR, we first introduce additional views for each image, where the texture undergoes significant alterations while preserving the image's underlying structure. We then encourage the model to learn the appearance-invariant representation by enforcing the consistency between object features across different views, for which we obtain class-agnostic object proposals using off-the-shelf unsupervised models that possess strong object-awareness. These proposals enable cross-view object feature matching, greatly reducing the appearance dependency while enhancing the object-awareness. We thoroughly evaluate our method on public benchmarks under both cross-class and cross-dataset settings, achieving state-of-the-art performance. Project page: https://visual-ai.github.io/vclr

Federated learning (FL) has emerged as a promising paradigm for privacy-preserving collaborative machine learning. However, extending FL to class incremental learning settings introduces three key challenges: 1) spatial heterogeneity due to non-IID data distributions across clients, 2) temporal heterogeneity due to sequential arrival of tasks, and 3) resource heterogeneity due to diverse client capabilities. Existing approaches generally address these challenges in isolation, potentially leading to interference between updates, catastrophic forgetting, or excessive communication overhead. In this paper, we propose personalized Federated class-incremental parameter efficient fine-tuning with Mixture of Frequency aggregation (pFedMixF), a novel framework that simultaneously addresses all three heterogeneity challenges through frequency domain decomposition. Our key insight is that assigning orthogonal frequency components to different clients and tasks enables interference-free learning to be achieved with minimal communication costs. We further design an Auto-Task Agnostic Classifier that automatically routes samples to task-specific classifiers while adapting to heterogeneous class distributions.We conduct extensive experiments on three benchmark datasets, comparing our approach with eight state-of-the-art methods. The results demonstrate that \methodname achieves comparable test accuracy while requiring only 25% of the entire model parameters and incurring significantly lower communication costs than baseline methods.

Semantic segmentation is a crucial prerequisite in clinical applications and computer-aided diagnosis. With the development of deep neural networks, biomedical image segmentation has achieved remarkable success. Encoder-decoder architectures that integrate convolutions and transformers are gaining attention for their potential to capture both global and local features. However, current designs face the contradiction that these two features cannot be continuously transmitted. In addition, some models lack a unified and standardized evaluation benchmark, leading to significant discrepancies in the experimental setup. In this study, we review and summarize these architectures and analyze their contradictions in design. We modify UNet and propose WNet to combine transformers and convolutions, addressing the transmission issue effectively. WNet captures long-range dependencies and local details simultaneously while ensuring their continuous transmission and multi-scale fusion. We integrate WNet into the nnUNet framework for unified benchmarking. Our model achieves state-of-the-art performance in biomedical image segmentation. Extensive experiments demonstrate their effectiveness on four 2D datasets (DRIVE, ISIC-2017, Kvasir-SEG, and CREMI) and four 3D datasets (Parse2022, AMOS22, BTCV, and ImageCAS). The code is available at https://github.com/Yanfeng-Zhou/nnWNet.

The efficient rendering and explicit nature of 3DGS promote the advancement of 3D scene manipulation.However, existing methods typically encounter challenges in controlling the manipulation region and are unable to furnish the user with interactive feedback, which inevitably leads to unexpected results.Intuitively, incorporating interactive 3D segmentation tools can compensate for this deficiency. Nevertheless, existing segmentation frameworks impose a pre-processing step of scene-specific parameter training, which limits the efficiency and flexibility of scene manipulation.To deliver a 3D region control module that is well-suited for scene manipulation with reliable efficiency, we propose **i**nteractive **Seg**ment-and-**Man**ipulate 3D Gaussians (**iSegMan**), an interactive segmentation and manipulation framework that only requires simple 2D user interactions in any view.To propagate user interactions to other views, we propose Epipolar-guided Interaction Propagation (**EIP**), which innovatively exploits epipolar constraint for efficient and robust interaction matching.To avoid scene-specific training to maintain efficiency, we further propose the novel Visibility-based Gaussian Voting (**VGV**), which obtains 2D segmentations from SAM and models the region extraction as a voting game between 2D Pixels and 3D Gaussians based on Gaussian visibility.Taking advantage of the efficient and precise region control of EIP and VGV, we put forth a **Manipulation Toolbox** to implement various functions on selected regions, enhancing the controllability, flexibility and practicality of scene manipulation.Extensive results on 3D scene manipulation and segmentation tasks fully demonstrate the significant advantages of iSegMan.

Traditional methods in pose estimation often rely on precise 3D models or additional data such as depth and normals, limiting their generalization, especially when objects undergo large translations or rotations. We propose iG-6DoF, a novel model-free 6D pose estimation method iterative 3D Gaussian Splatting to estimate the pose of unseen objects. We first estimates an initial pose by leveraging multi-scale data augmentation and the rotation-equivariant features to create a better pose hypothesis from a set of candidates. Then, we propose an iterative 3DGS approach through iteratively rendering and comparing the rendered image with the input image to further progressively improve pose estimation accuracy. The proposed end-to-end network consists of an object detector, a multi-scale rotation-equivariant feature based initial pose estimator, and a coarse-to-fine pose refiner. Such combination allows our method to focus on the target object in a complex scene and deal with large movement and weak textures.Such framework allows to deal with large movement and weak texture and complex scene. We conduct extensive experiments on benchmark and real datasets. Our method achieves state-of-the-art results on the LINEMOD, OnePose-LowTexture, and GenMOP datasets, demonstrating its strong generalization to unseen objects and robustness across various scenes.

We introduce a theoretical framework for diffusion-based image editing by formulating it as a reverse-time bridge modeling problem. This approach modifies the backward process of a pretrained diffusion model to construct a bridge that converges to an implicit distribution associated with the editing target at time 0. Building on this framework, we propose h-Edit, a novel editing method that utilizes Doob's h-transform and Langevin Monte Carlo to decompose the update of an intermediate edited sample into two components: a "reconstruction" term and an "editing" term. This decomposition provides flexibility, allowing the reconstruction term to be computed via existing inversion techniques and enabling the combination of multiple editing terms to handle complex editing tasks. To our knowledge, h-Edit is the first training-free method capable of performing simultaneous text-guided and reward-model-based editing. Extensive experiments, both quantitative and qualitative, show that h-Edit outperforms state-of-the-art baselines in terms of editing effectiveness and faithfulness.

We introduce Generalizable 3D-Language Feature Fields (g3D-LF), a 3D representation model pre-trained on large-scale 3D-language dataset for embodied tasks. Our g3D-LF processes posed RGB-D images from agents to encode feature fields for: 1) Novel view representation predictions from any position in the 3D scene; 2) Generations of BEV maps centered on the agent; 3) Querying targets using multi-granularity language within the above-mentioned representations. Our representation can be generalized to unseen environments, enabling real-time construction and dynamic updates. By volume rendering latent features along sampled rays and integrating semantic and spatial relationships through multiscale encoders, our g3D-LF produces representations at different scales and perspectives, aligned with multi-granularity language, via multi-level contrastive learning. Furthermore, we prepare a large-scale 3D-language dataset to align the representations of the feature fields with language. Extensive experiments on Vision-and-Language Navigation under both Panorama and Monocular settings, Zero-shot Object Navigation, and Situated Question Answering tasks highlight the significant advantages and effectiveness of our g3D-LF for embodied tasks. Our source code and dataset will be made open-source upon paper acceptance.

Federated learning has wide applications in the medical field. It enables knowledge sharing among different healthcare institutes while protecting patients' privacy. However, existing federated learning systems are typically centralized, requiring clients to upload client-specific knowledge to a central server for aggregation. This centralized approach would integrate the knowledge from each client into a centralized server, and the knowledge would be already undermined during the centralized integration before it reaches back to each client. Besides, the centralized approach also creates a dependency on the central server, which may affect training stability if the server malfunctions or connections are unstable. To address these issues, we propose a decentralized federated learning framework named dFLMoE. In our framework, clients directly exchange lightweight head models with each other. After exchanging, each client treats both local and received head models as individual experts, and utilizes a client-specific Mixture of Experts (MoE) approach to make collective decisions. This design not only reduces the knowledge damage with client-specific aggregations but also removes the dependency on the central server to enhance the robustness of the framework. We validate our framework on multiple medical tasks, demonstrating that our method evidently outperforms state-of-the-art approaches under both model homogeneity and heterogeneity settings.

Co-training has achieved significant success in the field of semi-supervised learning; however, the *homogenization phenomenon*, which arises from multiple models tending towards similar decision boundaries, remains inadequately addressed. To tackle this issue, we propose a novel algorithm called **β-FFT** from the perspectives of data diversity and training structure.First, from the perspective of data diversity, we introduce a nonlinear interpolation method based on the **Fast Fourier Transform (FFT)**. This method generates more diverse training samples by swapping low-frequency components between pairs of images, thereby enhancing the model's generalization capability. Second, from the structural perspective, we propose a differentiated training strategy to alleviate the homogenization issue in co-training. In this strategy, we apply additional training with labeled data to one model in the co-training framework, while employing linear interpolation based on the **Beta (β)** distribution for the unlabeled data as a regularization term for the additional training. This approach enables us to effectively utilize the limited labeled data while simultaneously improving the model's performance on unlabeled data, ultimately enhancing the overall performance of the system.Experimental results demonstrate that **β-FFT** outperforms current state-of-the-art (SOTA) methods on three public medical image datasets.

Diffusion models have revolutionized image generation, yet several challenges restrict their application to large-image domains, such as digital pathology and satellite imagery. Given that it is infeasible to directly train a model on 'whole' images from domains with potential gigapixel sizes, diffusion-based generative methods have focused on synthesizing small, fixed-size patches extracted from these images. However, generating small patches has limited applicability since patch-based models fail to capture the global structures and wider context of large images, which can be crucial for synthesizing (semantically) accurate samples. To overcome this limitation, we present ZoomLDM, a diffusion model tailored for generating images across multiple scales. Central to our approach is a novel magnification-aware conditioning mechanism that utilizes self-supervised learning (SSL) embeddings and allows the diffusion model to synthesize images at different 'zoom' levels, i.e., fixed-size patches extracted from large images at varying scales. ZoomLDM synthesizes coherent histopathology images that remain contextually accurate and detailed at different zoom levels, achieving state-of-the-art image generation quality across all scales and excelling in the data-scarce setting of generating thumbnails of entire large images. The multi-scale nature of ZoomLDM unlocks additional capabilities in large image generation, enabling computationally tractable and globally coherent image synthesis up to 4096x4096 pixels and 4x super-resolution. Additionally, multi-scale features extracted from ZoomLDM are highly effective in multiple instance learning experiments.

We introduce ZeroVO, a novel visual odometry (VO) algorithm that achieves zero-shot generalization across diverse cameras and environments, overcoming limitations in existing methods that depend on predefined or static camera calibration setups. Our approach incorporates three main innovations. First, we design a calibration-free, geometry-aware network structure capable of handling noise in estimated depth and camera parameters. Second, we introduce a language-based prior that infuses semantic information to enhance robust feature extraction and generalization to previously unseen domains. Third, we develop a flexible, semi-supervised training paradigm that iteratively adapts to new scenes using unlabeled data, further boosting the models' ability to generalize across diverse real-world scenarios. We analyze complex autonomous driving contexts, demonstrating over 30% improvement against prior methods on three standard benchmarks--KITTI, nuScenes, and Argoverse 2--as well as a newly introduced, high-fidelity synthetic dataset derived from Grand Theft Auto (GTA). By not requiring fine-tuning or camera calibration, our work broadens the applicability of VO, providing a versatile solution for real-world deployment at scale.

Robotic grasping is a cornerstone capability of embodied systems. Many methods directly output grasps from partial information without modeling the geometry of the scene, leading to suboptimal motion and even collisions. To address these issues, we introduce ZeroGrasp, a novel framework that simultaneously performs 3D reconstruction and grasp pose prediction in near real-time. A key insight of our method is that occlusion reasoning and modeling the spatial relationships between objects is beneficial for both accurate reconstruction and grasping. We couple our method with a novel large-scale synthetic dataset, which comprises 1M photo-realistic images, high-resolution 3D reconstructions and 11.3B physically-valid grasp pose annotations for 12K objects from the Objaverse-LVIS dataset. We evaluate ZeroGrasp on the GraspNet-1B benchmark as well as through real-world robot experiments. ZeroGrasp achieves state-of-the-art performance and generalizes to novel real-world objects by leveraging synthetic data.

This paper introduces ZeroMatch, a novel zero-shot RGB-D point cloud registration framework, aimed at achieving robust 3D matching on unseen data without any task-specific training. Our core idea is to utilize the powerful zero-shot image representation of Stable Diffusion, achieved through extensive pre-training on large-scale data, to enhance point-cloud geometric descriptors for robust matching. Specifically, we combine the handcrafted geometric descriptor FPFH with Stable-Diffusion features to create point descriptors that are both locally and contextually aware, enabling reliable RGB-D registration with zero-shot capability. This approach is based on our observation that Stable-Diffusion features effectively encode discriminative global contextual cues, naturally alleviating the feature ambiguity that FPFH often encounters in scenes with repetitive patterns or low overlap. To further enhance cross-view consistency of Stable-Diffusion features for improved matching, we propose a coupled-image input mode that concatenates the source and target images into a single input, replacing the original single-image mode. This design achieves both inter-image and prompt-to-image consistency attentions, facilitating robust cross-view feature interaction and alignment. Finally, we leverage feature nearest neighbors to construct putative correspondences for hypothesize-and-verify transformation estimation. Extensive experiments on 3DMatch, ScanNet, and ScanLoNet verify the excellent zero-shot matching ability of our method.

Recent advancements in 3D Large Multi-modal Models (3D-LMMs) have driven significant progress in 3D question answering. However, recent multi-frame Vision-Language Models (VLMs) demonstrate superior performance compared to 3D-LMMs on 3D question answering tasks, largely due to the greater scale and diversity of available 2D image data in contrast to the more limited 3D data. Multi-frame VLMs, although achieving superior performance, suffer from the difficulty of retaining all the detailed visual information in the 3D scene while limiting the number of visual tokens. Common methods such as token pooling, reduce visual token usage but often lead to information loss, impairing the model's ability to preserve visual details essential for 3D question answering tasks. To address this, we propose voxel-based Dynamic Token Compression (DTC), which combines 3D spatial priors and visual semantics to achieve over 90% reduction in visual tokens usage for current multi-frame VLMs. Our method maintains performance comparable to state-of-the-art models on 3D question answering benchmarks including OpenEQA and ScanQA, demonstrating its effectiveness.

Styled Handwritten Text Generation (HTG) has recently received attention from the computer vision and document analysis communities, which have developed several solutions, either GAN- or diffusion-based, that achieved promising results. Nonetheless, these strategies fail to generalize to novel styles and have technical constraints, particularly in terms of maximum output length and training efficiency. To overcome these limitations, in this work, we propose a novel framework for text-image generation, dubbed Emuru. Our approach leverages a powerful text-image representation model (a variational autoencoder) combined with an autoregressive Transformer. Our approach enables the generation of styled text images conditioned on textual content and style examples, such as specific fonts or handwriting styles. We train our model solely on a diverse, synthetic dataset of English text rendered in over 100,000 typewritten and calligraphy fonts, which gives it the capability to reproduce unseen styles (both fonts and users' handwriting) in zero-shot. To the best of our knowledge, Emuru is the first autoregressive model for HTG, and the first designed specifically for generalization to novel styles. Moreover, our model generates images without background artifacts, which are easier to use for downstream applications. Extensive evaluation on both typewritten and handwritten, any-length text image generation scenarios demonstrates the effectiveness of our approach.

Current methods for 3D scene reconstruction from sparse posed images employ intermediate 3D representations such as neural fields, voxel grids, or 3D Gaussians, to achieve multi-view consistent scene appearance and geometry. In this paper we introduce MVGD, a diffusion-based architecture capable of direct pixel-level generation of images and depth maps from novel viewpoints, given an arbitrary number of input views. Our method uses raymap conditioning to both augment visual features with spatial information from different viewpoints, as well as to guide the generation of images and depth maps from novel views. A key aspect of our approach is the multi-task generation of images and depth maps, using learnable task embeddings to guide the diffusion process towards specific modalities. We train this model on a collection of more than 60 million multi-view samples from publicly available datasets, and propose techniques to enable efficient and consistent learning in such diverse conditions. We also propose a novel strategy that enables the efficient training of larger models by incrementally fine-tuning smaller ones, with promising scaling behavior. Through extensive experiments, we report state-of-the-art results in multiple novel view synthesis benchmarks, as well as multi-view stereo and video depth estimation.

Large models have shown generalization across datasets for many low-level vision tasks, like depth estimation, but no such general models exist for scene flow.Even though scene flow prediction has wide potential, its practical use is limited because of the lack of generalization of current predictive models. We identify three key challenges and propose solutions for each. First, we create a method that jointly estimates geometry and motion for accurate prediction. Second, we alleviate scene flow data scarcity with a data recipe that affords us 1M annotated training samples across diverse synthetic scenes. Third, we evaluate different parameterizations for scene flow prediction and adopt a natural and effective parameterization. Our model outperforms existing methods as well as baselines built on large-scale models in terms of 3D end-point error, and shows zero-shot generalization to the casually captured videos from DAVIS and the robotic manipulation scenes from RoboTAP. Overall, our approach makes scene flow prediction more practical in-the-wild.

In recent years, it has become popular to tackle image restoration tasks with a single pretrained diffusion model (DM) and data-fidelity guidance, instead of training a dedicated deep neural network per task. However, such "zero-shot" restoration schemes currently require many Neural Function Evaluations (NFEs) for performing well, which may be attributed to the many NFEs needed in the original generative functionality of the DMs. Recently, faster variants of DMs have been explored for image generation. These include Consistency Models (CMs), which can generate samples via a couple of NFEs. However, existing works that use guided CMs for restoration still require tens of NFEs or fine-tuning of the model per task that leads to performance drop if the assumptions during the fine-tuning are not accurate. In this paper, we propose a zero-shot restoration scheme that uses CMs and operates well with as little as 4 NFEs. It is based on a wise combination of several ingredients: better initialization, back-projection guidance, and above all a novel noise injection mechanism. We demonstrate the advantages of our approach for image super-resolution and inpainting. Interestingly, we show that the usefulness of our noise injection technique goes beyond CMs: it can also mitigate the performance degradation of existing guided DM methods when reducing their NFE count.

With growing demand in media and social networks for personalized images, the need for advanced head-swapping techniques--integrating an entire head from the head image with the body from the body image--has increased. However, traditional head-swapping methods heavily rely on face-centered cropped data with primarily frontal-facing views, which limits their effectiveness in real-world applications. Additionally, their masking methods, designed to indicate regions requiring editing, are optimized for these types of dataset but struggle to achieve seamless blending in complex situations, such as when the original data includes features like long hair extending beyond the masked area. To overcome these limitations and enhance adaptability in diverse and complex scenarios, we propose a novel head swapping method, HID, that is robust to images including the full head and the upper body, and handles from frontal to side views, while automatically generating context-aware masks. For automatic mask generation, we introduce the IOMask, which enables seamless blending of the head and body, effectively addressing integration challenges. We further introduce the hair injection module to capture hair details with greater precision. Our experiments demonstrate that the proposed approach achieves state-of-the-art performance in head swapping, providing visually consistent and realistic results across a wide range of challenging conditions.