Vision-Language Models (VLMs) are crucial for real-world applications that require understanding textual and visual information. However, existing VLMs face multiple challenges in processing long videos, including computational inefficiency, memory limitations, and difficulties maintaining coherent understanding across extended sequences. These issues stem partly from the quadratic scaling of self-attention w.r.t. number of tokens but also encompass broader challenges in temporal reasoning and information integration over long sequences. To address these challenges, we introduce ReWind, a novel two-stage framework for long video understanding. In the first stage, ReWind maintains a dynamic memory that stores and updates instruction-relevant visual information as the video unfolds.Memory updates leverage novel read and write mechanisms utilizing learnable queries and cross-attentions between memory contents and the input stream. This approach maintains low memory requirements as the cross-attention layers scale linearly w.r.t. number of tokens. In the second stage, the memory content guides the selection of a few relevant frames, represented at high spatial resolution, which are combined with the memory contents and fed into an LLM to generate the final answer. We empirically demonstrate ReWind's superiority in visual question answering (VQA) and temporal grounding tasks, surpassing previous methods on long video benchmarks. Notably, ReWind achieves a +13% score gain and a +12% accuracy improvement on the MovieChat-1K VQA dataset and an +8% mIoU increase on Charades-STA for temporal grounding.
论文检索
输入标题、作者或关键词，从 2,871 篇学术成果中精准定位
Monocular Depth Estimation (MDE) has emerged as a pivotal task in computer vision, supporting numerous real-world applications. However, deploying accurate depth estimation models on resource-limited edge devices, especially Application-Specific Integrated Circuits (ASICs), is challenging due to the high computational and memory demands. Recent advancements in foundational depth estimation deliver impressive results but further amplify the difficulty of deployment on ASICs. To address this, we propose QuartDepth which adopts post-training quantization to quantize MDE models with hardware accelerations for ASICs. Our approach involves quantizing both weights and activations to 4-bit precision, reducing the model size and computation cost. To mitigate the performance degradation, we introduce activation polishing and compensation algorithm applied before and after activation quantization, as well as a weight reconstruction method for minimizing errors in weight quantization. Furthermore, we design a flexible and programmable hardware accelerator by supporting kernel fusion and customized instruction programmability, enhancing throughput and efficiency. Experimental results demonstrate that our framework achieves competitive accuracy while enabling fast inference and higher energy efficiency on ASICs, bridging the gap between high-performance depth estimation and practical edge-device applicability. Code: https://github.com/shawnricecake/quart-depth
Unlike global motion blur, Local Motion Deblurring (LMD) presents a more complex challenge, as it requires precise restoration of blurry regions while preserving the sharpness of the background. Existing LMD methods rely on manually annotated blur masks and often overlook the blur kernel's characteristics, which are crucial for accurate restoration. To address these limitations, we propose a novel parameterized motion kernel modeling approach that defines the motion blur kernel with three key parameters: length, angle, and curvature. We then use networks to estimate these kernel parameters, significantly improving the accuracy of blur kernel estimation. To effectively learn the motion blur representation, we incorporate a shared memory bank that stores blur prior information. Additionally, we introduce a dual-branch deblurring network: one branch leverages Mamba to capture long-range dependencies, while the other uses a mask-guided CNN focused on refining the local blurry regions. By fully utilizing the estimated bur prior information, our approach greatly enhances deblurring outcomes. Experimental results show that our method achieves state-of-the-art performance both quantitatively and visually, with a substantial reduction in computational complexity.
We investigate a new task in human motion prediction, which aims to forecast future body poses from historically observed sequences while accounting for arbitrary latency. This differs from existing works that assume an ideal scenario where future motions can be "instantaneously" predicted, thereby neglecting time delays caused by network transmission and algorithm execution. Addressing this task requires tackling two key challenges: The length of latency period can vary significantly across samples; the prediction model must be efficient. In this paper, we propose ALIEN, which treats the motion as a continuous function parameterized by a neural network, enabling predictions under any latency condition. By incorporating Mamba-like linear attention as a hyper-network and designing subsequent low-rank modulation, ALIEN efficiently learns a set of implicit neural representation weights from the observed motion to encode instance-specific information. Additionally, our model integrates the primary motion prediction task with an extra-designed variable-delay pose reconstruction task in a unified multi-task learning framework, enhancing its ability to capture richer motion patterns. Extensive experiments demonstrate that our approach outperforms state-of-the-art baselines adapted for our new task, while maintaining competitive performance in traditional prediction setting.
Beyond Local Sharpness: Communication-Efficient Global Sharpness-aware Minimization for Federated Learning
PDF ↗Federated learning (FL) enables collaborative model training with privacy preservation. Data heterogeneity across edge devices (clients) can cause models to converge to sharp minima, negatively impacting generalization and robustness. Recent approaches use client-side sharpness-aware minimization (SAM) to encourage flatter minima, but the discrepancy between local and global loss landscapes often undermines their effectiveness, as optimizing for local sharpness does not ensure global flatness. This work introduces FedGloSS (Federated Global Server-side Sharpness), a novel FL approach that prioritizes the optimization of global sharpness on the server, using SAM. To reduce communication overhead, FedGloSS cleverly approximates sharpness using the previous global gradient, eliminating the need for additional client communication. Our extensive evaluations demonstrate that FedGloSS consistently reaches flatter minima and better performance compared to state-of-the-art FL methods across various federated vision benchmarks. Code available at https://github.com/pietrocagnasso/fedgloss.
In this work, we introduce the Adventurer series models where we treat images as sequences of patch tokens and employ uni-directional language models to learn visual representations. This modeling paradigm allows us to process images in a recurrent formulation with linear complexity relative to the sequence length, which can effectively address the memory and computation explosion issues posed by high-resolution and fine-grained images. In detail, we introduce two simple designs that seamlessly integrate image inputs into the causal inference framework: a global pooling token placed at the beginning of the sequence and a flipping operation between every two layers. Extensive empirical studies highlight that compared with the existing plain architectures such as DeiT and Vim, Adventurer offers an optimal efficiency-accuracy trade-off. For example, our Adventurer-Base attains a competitive test accuracy of 84.3% on the standard ImageNet-1k benchmark with 216 images/s training throughput, which is 3.8x and 6.2x faster than Vim and DeiT to achieve the same result. As Adventurer offers great computation and memory efficiency and allows scaling with linear complexity, we hope this architecture can benefit future explorations in modeling long sequences for high-resolution or fine-grained images.
How can we enable models to comprehend video anomalies occurring over varying temporal scales and contexts?Traditional Video Anomaly Understanding (VAU) methods focus on frame-level anomaly prediction, often missing the interpretability of complex and diverse real-world anomalies. Recent multimodal approaches leverage visual and textual data but lack hierarchical annotations that capture both short-term and long-term anomalies.To address this challenge, we introduce HIVAU-70k, a large-scale benchmark for hierarchical video anomaly understanding across any granularity. We develop a semi-automated annotation engine that efficiently scales high-quality annotations by combining manual video segmentation with recursive free-text annotation using large language models (LLMs). This results in over 70,000 multi-granular annotations organized at clip-level, event-level, and video-level segments.For efficient anomaly detection in long videos, we propose the Anomaly-focused Temporal Sampler (ATS). ATS integrates an anomaly scorer with a density-aware sampler to adaptively select frames based on anomaly scores, ensuring that the multimodal LLM concentrates on anomaly-rich regions, which significantly enhances both efficiency and accuracy.Extensive experiments demonstrate that our hierarchical instruction data markedly improves anomaly comprehension. The integrated ATS and visual-language model outperform traditional methods in processing long videos.Our benchmark and model will be publicly available.
BOE-ViT: Boosting Orientation Estimation with Equivariance in Self-Supervised 3D Subtomogram Alignment
PDF ↗Subtomogram alignment is a critical task in cryo-electron tomography (cryo-ET) analysis, essential for achieving high-resolution reconstructions of macromolecular complexes. However, learning effective positional representations remains challenging due to limited labels and high noise levels inherent in cryo-ET data. In this work, we address this challenge by proposing a self-supervised learning approach that leverages intrinsic geometric transformations as implicit supervisory signals, enabling robust representation learning despite data scarcity. We introduce BOE-ViT, the first Vision Transformer (ViT) framework for 3D subtomogram alignment. Recognizing that traditional ViTs lack equivariance and are therefore suboptimal for orientation estimation, we enhance the model with two innovative modules that introduce equivariance include 1) the Polyshift module for improved shift estimation and 2) Multi-Axis Rotation Encoding (MARE) for enhanced rotation estimation. Experimental results demonstrate that BOE-ViT significantly outperforms state-of-the-art methods. Notably, at SNR 0.01 dataset, our approach achieves a 77.3% reduction in rotation estimation error and a 62.5% reduction in translation estimation error, effectively overcoming the challenges in cryo-ET subtomogram alignment.
Recent monocular 3D shape reconstruction methods have shown promising zero-shot results on object-segmented images without any occlusions. However, their effectiveness is significantly compromised in real-world conditions, due to imperfect object segmentation by off-the-shelf models and the prevalence of occlusions. To effectively address these issues, we propose a unified regression model that integrates segmentation and reconstruction, specifically designed for occlusion-aware 3D shape reconstruction. To facilitate its reconstruction in the wild, we also introduce a scalable data synthesis pipeline that simulates a wide range of variations in objects, occluders, and backgrounds. Training on our synthetic data enables the proposed model to achieve state-of-the-art zero-shot results on real-world images, using significantly fewer parameters than competing approaches.
USP-Gaussian: Unifying Spike-based Image Reconstruction, Pose Correction and Gaussian Splatting
PDF ↗Spike camera, as an innovative type of neuromorphic camera that captures scenes with 0-1 bit stream at 40 kHz, is increasingly being employed for the novel view synthesis task building on the techniques such as Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS). Previous spike-based approaches typically follow a three-stage pipeline: I. Spike-to-image reconstruction based on established algorithms. II. Camera poses estimation. III. Novel view synthesis. However, the cascading framework suffers from substantial cumulative errors, i.e., the quality of the initially reconstructed images will impact pose estimation, ultimately limiting the fidelity of the 3D reconstruction. To address this limitation, we propose a synergistic optimization framework USP-Gaussian, which unifies spike-to-image reconstruction, pose correction, and gaussian splatting into an end-to-end pipeline. Leveraging the multi-view consistency afforded by 3DGS and the motion capture capability of the spike camera, our framework enables iterative optimization between the spike-to-image reconstruction network and 3DGS. Experiments on synthetic datasets demonstrate that our method surpasses previous approaches by effectively eliminating cascading errors. Moreover, in real-world scenarios, our method achieves robust 3D reconstruction benefiting from the integration of pose optimization. Our code, data, and trained models are available at https://github.com/chenkang455/USP-Gaussian.
Developing a face anti-spoofing model that meets the security requirements of clients worldwide is challenging due to the domain gap between training datasets and the diverse end-user test data. Moreover, for security and privacy reasons, it is undesirable for clients to share a large amount of their face data with service providers. In this work, we introduce a novel method in which the face anti-spoofing model can be adapted by the client itself to a target domain at test time using only a small sample of data while keeping model parameters and training data inaccessible to the client. Specifically, we develop a prototype-based base model and an optimal transport-guided adaptor that enables adaptation in either a lightweight training or training-free fashion, without updating the base model's parameters. Furthermore, we propose geodesic mixup, an optimal transport-based synthesis method that generates augmented training data along the geodesic path between source prototypes and the target data distribution. This allows training a lightweight classifier to effectively adapt to target-specific characteristics while retaining essential knowledge learned from the source domain. In cross-domain and cross-attack settings, compared with recent methods, our method achieves average relative improvements of 19.17% in HTER and 8.58% in AUC, respectively.
3D Gaussian Splatting (3DGS) has emerged as a powerful technique for real-time, high-resolution novel view synthesis. By representing scenes as a mixture of Gaussian primitives, 3DGS leverages GPU rasterization pipelines for efficient rendering and reconstruction. To optimize scene coverage and capture fine details, 3DGS employs a densification algorithm to generate additional points. However, this process often leads to redundant point clouds, resulting in excessive memory usage, slower performance, and substantial storage demands - posing significant challenges for deployment on resource-constrained devices. To address this limitation, we propose a theoretical framework that demystifies and improves density control in 3DGS. Our analysis reveals that splitting is crucial for escaping saddle points. Through an optimization-theoretic approach, we establish the necessary conditions for densification, determine the minimal number of offspring Gaussians, identify the optimal parameter update direction, and provide an analytical solution for normalizing off-spring opacity. Building on these insights, we introduce SteepGS, incorporating steepest density control, a principled strategy that minimizes loss while maintaining a compact point cloud. SteepGS achieves a 50% reduction in Gaussian points without compromising rendering quality, significantly enhancing both efficiency and scalability.
High-fidelity 3D Object Generation from Single Image with RGBN-Volume Gaussian Reconstruction Model
PDF ↗Recently single-view 3D generation via Gaussian splatting has emerged and developed quickly. They learn 3D Gaussians from 2D RGB images generated from pre-trained multi-view diffusion (MVD) models, and have shown a promising avenue for 3D generation through a single image. Despite the current progress, these methods still suffer from the inconsistency jointly caused by the geometric ambiguity in the 2D images, and the lack of structure of 3D Gaussians, leading to distorted and blurry 3D object generation. In this paper, we propose to fix these issues by GS-RGBN, a new RGBN-volume Gaussian Reconstruction Model designed to generate high-fidelity 3D objects from single-view images. Our key insight is a structured 3D representation can simultaneously mitigate the afore-mentioned two issues. To this end, we propose a novel hybrid Voxel-Gaussian representation, where a 3D voxel representation contains explicit 3D geometric information, eliminating the geometric ambiguity from 2D images. It also structures Gaussians during learning so that the optimization tends to find better local optima. Our 3D voxel representation is obtained by a fusion module that aligns RGB features and surface normal features, both of which can be estimated from 2D images. Extensive experiments demonstrate the superiority of our methods over prior works in terms of high-quality reconstruction results, robust generalization, and good efficiency.
SSHNet: Unsupervised Cross-modal Homography Estimation via Problem Reformulation and Split Optimization
PDF ↗We propose a novel unsupervised cross-modal homography estimation learning framework, named Split Supervised Homography estimation Network (SSHNet). SSHNet reformulates the unsupervised cross-modal homography estimation into two supervised sub-problems, each addressed by its specialized network: a homography estimation network and a modality transfer network. To realize stable training, we introduce an effective split optimization strategy to train each network separately within its respective sub-problem. We also formulate an extra homography feature space supervision to enhance feature consistency, further boosting the estimation accuracy. Moreover, we employ a simple yet effective distillation training technique to reduce model parameters and improve cross-domain generalization ability while maintaining comparable performance. The training stability of SSHNet enables its cooperation with various homography estimation architectures. Experiments reveal that the SSHNet using IHN as homography estimation network, namely SSHNet-IHN, outperforms previous unsupervised approaches by a significant margin. Even compared to supervised approaches MHN and LocalTrans, SSHNet-IHN achieves 47.4% and 85.8% mean average corner errors (MACEs) reduction on the challenging OPT-SAR dataset. Source code is available at https://github.com/Junchen-Yu/SSHNet.
GeoDepth: From Point-to-Depth to Plane-to-Depth Modeling for Self-Supervised Monocular Depth Estimation
PDF ↗Self-supervised monocular depth estimation has long been treated as a point-wise prediction problem, where the depth of each pixel is usually estimated independently. However, artifacts are often observed in the estimated depth map, e.g., depth values for points located in the same region may jump dramatically. To address this issue, we propose a novel self-supervised monocular depth estimation framework called GeoDepth, where we explore the intrinsic geometric representation in 3D scenes for producing accurate and continuous depth maps. In particular, we model the complex 3D scene as a collection of planes with varying sizes, where each plane is characterized by a unique set of parameters, namely planar normal (indicating plane orientation) and planar offset (defining the perpendicular distance from the camera center to the plane). Under this modeling, points in the same plane are enforced to share a unique representation and their depth variations related only to pixel coordinates, thus this geometric relationship can be exploited to regularize the depth variations of these points. To this end, we design a structured plane generation module that introduces spatio-temporal geometric cues and the plane uniqueness principle to recover the correct scene plane representation. In addition, we develop a depth discontinuity module to identify depth discontinuity regions and subsequently optimize them. Our experiments on the KITTI and NYUv2 datasets demonstrate that GeoDepth achieves state-of-the-art performance, with additional tests on Make3D and ScanNet validating its generalization capabilities.
Upcycled Mixture-of-Experts (MoE) models have shown great potential in various tasks by converting the original Feed-Forward Network (FFN) layers in pre-trained dense models into MoE layers. However, these models still suffer from significant parameter inefficiency due to the introduction of multiple experts. In this work, we propose a novel DeRS (Decompose, Replace, and Synthesis) paradigm to overcome this shortcoming, which is motivated by our observations about the unique redundancy mechanisms of upcycled MoE experts. Specifically, DeRS decomposes the experts into one expert-shared base weight and multiple expert-specific delta weights, and subsequently represents these delta weights in lightweight forms. Our proposed DeRS paradigm can be applied to enhance parameter efficiency in two different scenarios, including: 1) DeRS Compression for inference stage, using sparsification or quantization to compress vanilla upcycled MoE models; and 2) DeRS Upcycling for training stage, employing lightweight sparse or low-rank matrixes to efficiently upcycle dense models into MoE models. Extensive experiments across three different tasks show that the proposed methods can achieve extreme parameter efficiency while maintaining the performance for both training and compression of upcycled MoE models.
Video semantic segmentation (VSS) plays a vital role in understanding the temporal evolution of scenes. Traditional methods often segment videos frame-by-frame or in a short temporal window, leading to limited temporal context, redundant computations, and heavy memory requirements. To this end, we introduce a Temporal Video State Space Sharing (TV3S) architecture to leverage Mamba state space models for temporal feature sharing. Our model features a selective gating mechanism that efficiently propagates relevant information across video frames, eliminating the need for a memory-heavy feature pool. By processing spatial patches independently and incorporating shifted operation, TV3S supports highly parallel computation in both training and inference stages, which reduces the delay in sequential state space processing and improves the scalability for long video sequences. Moreover, TV3S incorporates information from prior frames during inference, achieving long-range temporal coherence and superior adaptability to extended sequences. Evaluations on the VSPW and Cityscapes datasets reveal that our approach outperforms current state-of-the-art methods, establishing a new standard for VSS with consistent results across long video sequences. By achieving a good balance between accuracy and efficiency, TV3S shows a significant advancement in spatiotemporal modeling, paving the way for efficient video analysis. The code is publicly available at https://github.com/Ashesham/TV3S.git.
This work introduces ILIAS, a new test dataset for Instance-Level Image retrieval At Scale. It is designed to evaluate the ability of current and future foundation models and retrieval techniques to recognize particular objects. The key benefits over existing datasets include large scale, domain diversity, accurate ground truth, and a performance that is far from saturated. ILIAS includes query and positive images for 1,000 object instances, manually collected to capture challenging conditions and diverse domains. Large-scale retrieval is conducted against 100 million distractor images from YFCC100M. To avoid false negatives without extra annotation effort, we include only query objects confirmed to have emerged after 2014, i.e. the compilation date of YFCC100M. An extensive benchmarking is performed with the following observations: i) models fine-tuned on specific domains, such as landmarks or products, excel in that domain but fail on ILIAS ii) learning a linear adaptation layer using multi-domain class supervision results in performance improvements, especially for vision-language models iii) local descriptors in retrieval re-ranking are still a key ingredient, especially in the presence of severe background clutter iv) the text-to-image performance of the vision-language foundation models is surprisingly close to the corresponding image-to-image case. website: https://vrg.fel.cvut.cz/ilias/
Existing unlearning algorithms in text-to-image generative models often fail to preserve the knowledge of semantically related concepts when removing specific target concepts--a challenge known as adjacency. To address this, we propose FADE (Fine-grained Attenuation for Diffusion Erasure), introducing adjacency-aware unlearning in diffusion models. FADE comprises two components: (1) the Concept Neighborhood, which identifies an adjacency set of related concepts, and (2) Mesh Modules, employing a structured combination of Expungement, Adjacency, and Guidance loss components. These enable precise erasure of target concepts while preserving fidelity across related and unrelated concepts. Evaluated on datasets like Stanford Dogs, Oxford Flowers, CUB, I2P, Imagenette, and ImageNet-1k, FADE effectively removes target concepts with minimal impact on correlated concepts, achieving at least a 12% improvement in retention performance over state-of-the-art methods. Our code and models are available on the project page: iab-rubric/unlearning/FG-Un.
All-in-one models for adverse weather removal aim to process various degraded images using a single set of parameters, making them ideal for real-world scenarios. However, they encounter two main challenges: catastrophic forgetting and limited degradation awareness. The former causes the model to lose knowledge of previously learned scenarios, reducing its overall effectiveness. While the later hampers the model's ability to accurately identify and respond to specific types of degradation, limiting its performance across diverse adverse weather conditions. To address these issues, we introduce the Incremental Learning Adverse Weather Removal (ILAWR) framework, which uses a novel degradation-aware distillation strategy for continuous weather removal. Specifically, we first design a degradation-aware module that utilizes Fourier priors to capture a broad range of degradation features, effectively mitigating catastrophic forgetting in low-level visual tasks. Then, we implement multilateral distillation, which combines knowledge from multiple teacher models using an importance-guided aggregation approach. This enables the model to balance adaptation to new degradation types with the preservation of background details. Extensive experiments on both synthetic and real-world datasets confirm that ILAWR outperforms existing models across multiple benchmarks, proving its effectiveness in continuous adverse weather removal.