arXiv subjects

Yunsong Li

Publications and source records attributed to Yunsong Li.

At least 19 recordsLinked to original sources

arXiv2026

MobileSAM2: Lightweight Segment Anything for Spatial Intelligence

The recent large video foundation model, SAM2, enables segment anything in both images and videos, serving as a powerful base model for various applications. However, many of such use cases require to operate on resource-constrained devices like mobile phones and laptops. In this work, we aim to make SAM2 more mobile-friendly by distilling the heavyweight SAM2 into a lightweight model, facilitating segment anything in both images and videos on mobile devices. To this end, we propose Hypergraphical Knowledge Distill (HyperKD), which introduces the idea of hypergraph into knowledge distillation, aiming to effectively model and transfer SAM2's generalizable and comprehensive knowledge. HyperKD consists of Temporal HyperKD and Granularity HyperKD that construct hypergraphs to explicitly model and extract the generalizable temporal knowledge and the comprehensive multi-granularity knowledge from SAM2 respectively, which are then distilled into the lightweight student model by aligning it with the constructed hypergraphs. Besides, we present MobileSAM2, a new family of lightweight SAM2 that balances efficiency and effectiveness via searching the best model architectures with HyperKD during model size reduction. Extensive experiments validate MobileSAM2 across multiple benchmarks and show promising generalization performance on embodied AI tasks.

cs.CV↗

arXiv2025

High-Energy Concentration for Federated Learning in Frequency Domain

Federated Learning (FL) presents significant potential for collaborative optimization without data sharing. Since synthetic data is sent to the server, leveraging the popular concept of dataset distillation, this FL framework protects real data privacy while alleviating data heterogeneity. However, such methods are still challenged by the redundant information and noise in entire spatial-domain designs, which inevitably increases the communication burden. In this paper, we propose a novel Frequency-Domain aware FL method with high-energy concentration (FedFD) to address this problem. Our FedFD is inspired by the discovery that the discrete cosine transform predominantly distributes energy to specific regions, referred to as high-energy concentration. The principle behind FedFD is that low-energy like high-frequency components usually contain redundant information and noise, thus filtering them helps reduce communication costs and optimize performance. Our FedFD is mathematically formulated to preserve the low-frequency components using a binary mask, facilitating an optimal solution through frequency-domain distribution alignment. In particular, real data-driven synthetic classification is imposed into the loss to enhance the quality of the low-frequency components. On five image and speech datasets, FedFD achieves superior performance than state-of-the-art methods while reducing communication costs. For example, on the CIFAR-10 dataset with Dirichlet coefficient $α= 0.01$, FedFD achieves a minimum reduction of 37.78\% in the communication cost, while attaining a 10.88\% performance gain.

cs.LG↗

arXiv2025

Hyperspectral Mamba for Hyperspectral Object Tracking

Hyperspectral object tracking holds great promise due to the rich spectral information and fine-grained material distinctions in hyperspectral images, which are beneficial in challenging scenarios. While existing hyperspectral trackers have made progress by either transforming hyperspectral data into false-color images or incorporating modality fusion strategies, they often fail to capture the intrinsic spectral information, temporal dependencies, and cross-depth interactions. To address these limitations, a new hyperspectral object tracking network equipped with Mamba (HyMamba), is proposed. It unifies spectral, cross-depth, and temporal modeling through state space modules (SSMs). The core of HyMamba lies in the Spectral State Integration (SSI) module, which enables progressive refinement and propagation of spectral features with cross-depth and temporal spectral information. Embedded within each SSI, the Hyperspectral Mamba (HSM) module is introduced to learn spatial and spectral information synchronously via three directional scanning SSMs. Based on SSI and HSM, HyMamba constructs joint features from false-color and hyperspectral inputs, and enhances them through interaction with original spectral features extracted from raw hyperspectral images. Extensive experiments conducted on seven benchmark datasets demonstrate that HyMamba achieves state-of-the-art performance. For instance, it achieves 73.0\% of the AUC score and 96.3\% of the DP@20 score on the HOTC2020 dataset. The code will be released at https://github.com/lgao001/HyMamba.

cs.CV↗

arXiv2025

FusionSAM: Visual Multi-Modal Learning with Segment Anything

Multimodal image fusion and semantic segmentation are critical for autonomous driving. Despite advancements, current models often struggle with segmenting densely packed elements due to a lack of comprehensive fusion features for guidance during training. While the Segment Anything Model (SAM) allows precise control during fine-tuning through its flexible prompting encoder, its potential remains largely unexplored in the context of multimodal segmentation for natural images. In this paper, we introduce SAM into multimodal image segmentation for the first time, proposing a novel framework that combines Latent Space Token Generation (LSTG) and Fusion Mask Prompting (FMP) modules. This approach transforms the training methodology for multimodal segmentation from a traditional black-box approach to a controllable, prompt-based mechanism. Specifically, we obtain latent space features for both modalities through vector quantization and embed them into a cross-attention-based inter-domain fusion module to establish long-range dependencies between modalities. We then use these comprehensive fusion features as prompts to guide precise pixel-level segmentation. Extensive experiments on multiple public datasets demonstrate that our method significantly outperforms SAM and SAM2 in multimodal autonomous driving scenarios, achieving an average improvement of 4.1$\%$ over the state-of-the-art method in segmentation mIoU, and the performance is also optimized in other multi-modal visual scenes.

cs.CV↗

arXiv2025

Hyperspectral Adapter for Object Tracking based on Hyperspectral Video

Object tracking based on hyperspectral video attracts increasing attention to the rich material and motion information in the hyperspectral videos. The prevailing hyperspectral methods adapt pretrained RGB-based object tracking networks for hyperspectral tasks by fine-tuning the entire network on hyperspectral datasets, which achieves impressive results in challenging scenarios. However, the performance of hyperspectral trackers is limited by the loss of spectral information during the transformation, and fine-tuning the entire pretrained network is inefficient for practical applications. To address the issues, a new hyperspectral object tracking method, hyperspectral adapter for tracking (HyA-T), is proposed in this work. The hyperspectral adapter for the self-attention (HAS) and the hyperspectral adapter for the multilayer perceptron (HAM) are proposed to generate the adaption information and to transfer the multi-head self-attention (MSA) module and the multilayer perceptron (MLP) in pretrained network for the hyperspectral object tracking task by augmenting the adaption information into the calculation of the MSA and MLP. Additionally, the hyperspectral enhancement of input (HEI) is proposed to augment the original spectral information into the input of the tracking network. The proposed methods extract spectral information directly from the hyperspectral images, which prevent the loss of the spectral information. Moreover, only the parameters in the proposed methods are fine-tuned, which is more efficient than the existing methods. Extensive experiments were conducted on four datasets with various spectral bands, verifing the effectiveness of the proposed methods. The HyA-T achieves state-of-the-art performance on all the datasets.

cs.CV↗

arXiv2025

M$^3$amba: CLIP-driven Mamba Model for Multi-modal Remote Sensing Classification

Multi-modal fusion holds great promise for integrating information from different modalities. However, due to a lack of consideration for modal consistency, existing multi-modal fusion methods in the field of remote sensing still face challenges of incomplete semantic information and low computational efficiency in their fusion designs. Inspired by the observation that the visual language pre-training model CLIP can effectively extract strong semantic information from visual features, we propose M$^3$amba, a novel end-to-end CLIP-driven Mamba model for multi-modal fusion to address these challenges. Specifically, we introduce CLIP-driven modality-specific adapters in the fusion architecture to avoid the bias of understanding specific domains caused by direct inference, making the original CLIP encoder modality-specific perception. This unified framework enables minimal training to achieve a comprehensive semantic understanding of different modalities, thereby guiding cross-modal feature fusion. To further enhance the consistent association between modality mappings, a multi-modal Mamba fusion architecture with linear complexity and a cross-attention module Cross-SS2D are designed, which fully considers effective and efficient information interaction to achieve complete fusion. Extensive experiments have shown that M$^3$amba has an average performance improvement of at least 5.98\% compared with the state-of-the-art methods in multi-modal hyperspectral image classification tasks in the remote sensing field, while also demonstrating excellent training efficiency, achieving a double improvement in accuracy and efficiency. The code is released at https://github.com/kaka-Cao/M3amba.

cs.CV↗

arXiv2025

E2E-MFD: Towards End-to-End Synchronous Multimodal Fusion Detection

Multimodal image fusion and object detection are crucial for autonomous driving. While current methods have advanced the fusion of texture details and semantic information, their complex training processes hinder broader applications. Addressing this challenge, we introduce E2E-MFD, a novel end-to-end algorithm for multimodal fusion detection. E2E-MFD streamlines the process, achieving high performance with a single training phase. It employs synchronous joint optimization across components to avoid suboptimal solutions tied to individual tasks. Furthermore, it implements a comprehensive optimization strategy in the gradient matrix for shared parameters, ensuring convergence to an optimal fusion detection configuration. Our extensive testing on multiple public datasets reveals E2E-MFD's superior capabilities, showcasing not only visually appealing image fusion but also impressive detection outcomes, such as a 3.9% and 2.0% mAP50 increase on horizontal object detection dataset M3FD and oriented object detection dataset DroneVehicle, respectively, compared to state-of-the-art approaches. The code is released at https://github.com/icey-zhang/E2E-MFD.

cs.CV↗

arXiv2024

DiffCLIP: Few-shot Language-driven Multimodal Classifier

Visual language models like Contrastive Language-Image Pretraining (CLIP) have shown impressive performance in analyzing natural images with language information. However, these models often encounter challenges when applied to specialized domains such as remote sensing due to the limited availability of image-text pairs for training. To tackle this issue, we introduce DiffCLIP, a novel framework that extends CLIP to effectively convey comprehensive language-driven semantic information for accurate classification of high-dimensional multimodal remote sensing images. DiffCLIP is a few-shot learning method that leverages unlabeled images for pretraining. It employs unsupervised mask diffusion learning to capture the distribution of diverse modalities without requiring labels. The modality-shared image encoder maps multimodal data into a unified subspace, extracting shared features with consistent parameters across modalities. A well-trained image encoder further enhances learning by aligning visual representations with class-label text information from CLIP. By integrating these approaches, DiffCLIP significantly boosts CLIP performance using a minimal number of image-text pairs. We evaluate DiffCLIP on widely used high-dimensional multimodal datasets, demonstrating its effectiveness in addressing few-shot annotated classification tasks. DiffCLIP achieves an overall accuracy improvement of 10.65% across three remote sensing datasets compared with CLIP, while utilizing only 2-shot image-text pairs. The code has been released at https://github.com/icey-zhang/DiffCLIP.

cs.CV↗

arXiv2024

Towards Accurate and Efficient Sub-8-Bit Integer Training

Neural network training is a memory- and compute-intensive task. Quantization, which enables low-bitwidth formats in training, can significantly mitigate the workload. To reduce quantization error, recent methods have developed new data formats and additional pre-processing operations on quantizers. However, it remains quite challenging to achieve high accuracy and efficiency simultaneously. In this paper, we explore sub-8-bit integer training from its essence of gradient descent optimization. Our integer training framework includes two components: ShiftQuant to realize accurate gradient estimation, and L1 normalization to smoothen the loss landscape. ShiftQuant attains performance that approaches the theoretical upper bound of group quantization. Furthermore, it liberates group quantization from inefficient memory rearrangement. The L1 normalization facilitates the implementation of fully quantized normalization layers with impressive convergence accuracy. Our method frees sub-8-bit integer training from pre-processing and supports general devices. This framework achieves negligible accuracy loss across various neural networks and tasks ($0.92\%$ on 4-bit ResNets, $0.61\%$ on 6-bit Transformers). The prototypical implementation of ShiftQuant achieves more than $1.85\times/15.3\%$ performance improvement on CPU/GPU compared to its FP16 counterparts, and $33.9\%$ resource consumption reduction on FPGA than the FP16 counterparts. The proposed fully-quantized L1 normalization layers achieve more than $35.54\%$ improvement in throughout on CPU compared to traditional L2 normalization layers. Moreover, theoretical analysis verifies the advancement of our method.

cs.LG↗

arXiv2024

An End-to-End Real-World Camera Imaging Pipeline

Recent advances in neural camera imaging pipelines have demonstrated notable progress. Nevertheless, the real-world imaging pipeline still faces challenges including the lack of joint optimization in system components, computational redundancies, and optical distortions such as lens shading.In light of this, we propose an end-to-end camera imaging pipeline (RealCamNet) to enhance real-world camera imaging performance. Our methodology diverges from conventional, fragmented multi-stage image signal processing towards end-to-end architecture. This architecture facilitates joint optimization across the full pipeline and the restoration of coordinate-biased distortions. RealCamNet is designed for high-quality conversion from RAW to RGB and compact image compression. Specifically, we deeply analyze coordinate-dependent optical distortions, e.g., vignetting and dark shading, and design a novel Coordinate-Aware Distortion Restoration (CADR) module to restore coordinate-biased distortions. Furthermore, we propose a Coordinate-Independent Mapping Compression (CIMC) module to implement tone mapping and redundant information compression. Existing datasets suffer from misalignment and overly idealized conditions, making them inadequate for training real-world imaging pipelines. Therefore, we collected a real-world imaging dataset. Experiment results show that RealCamNet achieves the best rate-distortion performance with lower inference latency.

eess.IV↗

arXiv2024

Dual Inverse Degradation Network for Real-World SDRTV-to-HDRTV Conversion

In this study, we address the emerging necessity of converting Standard Dynamic Range Television (SDRTV) content into High Dynamic Range Television (HDRTV) in light of the limited number of native HDRTV content. A principal technical challenge in this conversion is the exacerbation of coding artifacts inherent in SDRTV, which detrimentally impacts the quality of the resulting HDRTV. To address this issue, our method introduces a novel approach that conceptualizes the SDRTV-to-HDRTV conversion as a composite task involving dual degradation restoration. This encompasses inverse tone mapping in conjunction with video restoration. We propose Dual Inversion Downgraded SDRTV to HDRTV Network (DIDNet), which can accurately perform inverse tone mapping while preventing encoding artifacts from being amplified, thereby significantly improving visual quality. DIDNet integrates an intermediate auxiliary loss function to effectively separate the dual degradation restoration tasks and efficient learning of both artifact reduction and inverse tone mapping during end-to-end training. Additionally, DIDNet introduces a spatio-temporal feature alignment module for video frame fusion, which augments texture quality and reduces artifacts. The architecture further includes a dual-modulation convolution mechanism for optimized inverse tone mapping. Recognizing the richer texture and high-frequency information in HDRTV compared to SDRTV, we further introduce a wavelet attention module to enhance frequency features. Our approach demonstrates marked superiority over existing state-of-the-art techniques in terms of quantitative performance and visual quality.

eess.IV↗

arXiv2024

SeaDATE: Remedy Dual-Attention Transformer with Semantic Alignment via Contrast Learning for Multimodal Object Detection

Multimodal object detection leverages diverse modal information to enhance the accuracy and robustness of detectors. By learning long-term dependencies, Transformer can effectively integrate multimodal features in the feature extraction stage, which greatly improves the performance of multimodal object detection. However, current methods merely stack Transformer-guided fusion techniques without exploring their capability to extract features at various depth layers of network, thus limiting the improvements in detection performance. In this paper, we introduce an accurate and efficient object detection method named SeaDATE. Initially, we propose a novel dual attention Feature Fusion (DTF) module that, under Transformer's guidance, integrates local and global information through a dual attention mechanism, strengthening the fusion of modal features from orthogonal perspectives using spatial and channel tokens. Meanwhile, our theoretical analysis and empirical validation demonstrate that the Transformer-guided fusion method, treating images as sequences of pixels for fusion, performs better on shallow features' detail information compared to deep semantic information. To address this, we designed a contrastive learning (CL) module aimed at learning features of multimodal samples, remedying the shortcomings of Transformer-guided fusion in extracting deep semantic features, and effectively utilizing cross-modal information. Extensive experiments and ablation studies on the FLIR, LLVIP, and M3FD datasets have proven our method to be effective, achieving state-of-the-art detection performance.

cs.CV↗

arXiv2024

Multi-scale direction-aware SAR object detection network via global information fusion

Deep learning has driven significant progress in object detection using Synthetic Aperture Radar (SAR) imagery. Existing methods, while achieving promising results, often struggle to effectively integrate local and global information, particularly direction-aware features. This paper proposes SAR-Net, a novel framework specifically designed for global fusion of direction-aware information in SAR object detection. SAR-Net leverages two key innovations: the Unity Compensation Mechanism (UCM) and the Direction-aware Attention Module (DAM). UCM facilitates the establishment of complementary relationships among features across different scales, enabling efficient global information fusion and transmission. Additionally, DAM, through bidirectional attention polymerization, captures direction-aware information, effectively eliminating background interference. Extensive experiments demonstrate the effectiveness of SAR-Net, achieving state-of-the-art results on aircraft (SAR-AIRcraft-1.0) and ship datasets (SSDD, HRSID), confirming its generalization capability and robustness.

cs.CV↗

arXiv2024

AMPO: Automatic Multi-Branched Prompt Optimization

Prompt engineering is very important to enhance the performance of large language models (LLMs). When dealing with complex issues, prompt engineers tend to distill multiple patterns from examples and inject relevant solutions to optimize the prompts, achieving satisfying results. However, existing automatic prompt optimization techniques are only limited to producing single flow instructions, struggling with handling diverse patterns. In this paper, we present AMPO, an automatic prompt optimization method that can iteratively develop a multi-branched prompt using failure cases as feedback. Our goal is to explore a novel way of structuring prompts with multi-branches to better handle multiple patterns in complex tasks, for which we introduce three modules: Pattern Recognition, Branch Adjustment, and Branch Pruning. In experiments across five tasks, AMPO consistently achieves the best results. Additionally, our approach demonstrates significant optimization efficiency due to our adoption of a minimal search strategy.

cs.CL↗

arXiv2024

Unleashing the Power of Generic Segmentation Models: A Simple Baseline for Infrared Small Target Detection

Recent advancements in deep learning have greatly advanced the field of infrared small object detection (IRSTD). Despite their remarkable success, a notable gap persists between these IRSTD methods and generic segmentation approaches in natural image domains. This gap primarily arises from the significant modality differences and the limited availability of infrared data. In this study, we aim to bridge this divergence by investigating the adaptation of generic segmentation models, such as the Segment Anything Model (SAM), to IRSTD tasks. Our investigation reveals that many generic segmentation models can achieve comparable performance to state-of-the-art IRSTD methods. However, their full potential in IRSTD remains untapped. To address this, we propose a simple, lightweight, yet effective baseline model for segmenting small infrared objects. Through appropriate distillation strategies, we empower smaller student models to outperform state-of-the-art methods, even surpassing fine-tuned teacher results. Furthermore, we enhance the model's performance by introducing a novel query design comprising dense and sparse queries to effectively encode multi-scale features. Through extensive experimentation across four popular IRSTD datasets, our model demonstrates significantly improved performance in both accuracy and throughput compared to existing approaches, surpassing SAM and Semantic-SAM by over 14 IoU on NUDT and 4 IoU on IRSTD1k. The source code and models will be released at https://github.com/O937-blip/SimIR.

cs.CV↗

arXiv2024

FedFQ: Federated Learning with Fine-Grained Quantization

Federated learning (FL) is a decentralized approach, enabling multiple participants to collaboratively train a model while ensuring the protection of data privacy. The transmission of updates from numerous edge clusters to the server creates a significant communication bottleneck in FL. Quantization is an effective compression technology, showcasing immense potential in addressing this bottleneck problem. The Non-IID nature of FL renders it sensitive to quantization. Existing quantized FL frameworks inadequately balance high compression ratios and superior convergence performance by roughly employing a uniform quantization bit-width on the client-side. In this work, we propose a communication-efficient FL algorithm with a fine-grained adaptive quantization strategy (FedFQ). FedFQ addresses the trade-off between achieving high communication compression ratios and maintaining superior convergence performance by introducing parameter-level quantization. Specifically, we have designed a Constraint-Guided Simulated Annealing algorithm to determine specific quantization schemes. We derive the convergence of FedFQ, demonstrating its superior convergence performance compared to existing quantized FL algorithms. We conducted extensive experiments on multiple benchmarks and demonstrated that, while maintaining lossless performance, FedFQ achieves a compression ratio of 27 times to 63 times compared to the baseline experiment.

cs.DC↗

arXiv2024

DA-BEV: Unsupervised Domain Adaptation for Bird's Eye View Perception

Camera-only Bird's Eye View (BEV) has demonstrated great potential in environment perception in a 3D space. However, most existing studies were conducted under a supervised setup which cannot scale well while handling various new data. Unsupervised domain adaptive BEV, which effective learning from various unlabelled target data, is far under-explored. In this work, we design DA-BEV, the first domain adaptive camera-only BEV framework that addresses domain adaptive BEV challenges by exploiting the complementary nature of image-view features and BEV features. DA-BEV introduces the idea of query into the domain adaptation framework to derive useful information from image-view and BEV features. It consists of two query-based designs, namely, query-based adversarial learning (QAL) and query-based self-training (QST), which exploits image-view features or BEV features to regularize the adaptation of the other. Extensive experiments show that DA-BEV achieves superior domain adaptive BEV perception performance consistently across multiple datasets and tasks such as 3D object detection and 3D scene segmentation.

cs.CV↗

arXiv2024

Reducing Spurious Correlation for Federated Domain Generalization

The rapid development of multimedia has provided a large amount of data with different distributions for visual tasks, forming different domains. Federated Learning (FL) can efficiently use this diverse data distributed on different client media in a decentralized manner through model sharing. However, in open-world scenarios, there is a challenge: global models may struggle to predict well on entirely new domain data captured by certain media, which were not encountered during training. Existing methods still rely on strong statistical correlations between samples and labels to address this issue, which can be misleading, as some features may establish spurious short-cut correlations with the predictions. To comprehensively address this challenge, we introduce FedCD (Cross-Domain Invariant Federated Learning), an overall optimization framework at both the local and global levels. We introduce the Spurious Correlation Intervener (SCI), which employs invariance theory to locally generate interventers for features in a self-supervised manner to reduce the model's susceptibility to spurious correlated features. Our approach requires no sharing of data or features, only the gradients related to the model. Additionally, we develop the simple yet effective Risk Extrapolation Aggregation strategy (REA), determining aggregation coefficients through mathematical optimization to facilitate global causal invariant predictions. Extensive experiments and ablation studies highlight the effectiveness of our approach. In both classification and object detection generalization tasks, our method outperforms the baselines by an average of at least 1.45% in Acc, 4.8% and 1.27% in mAP50.

cs.CV↗