arXiv subjects

Shengyu Li

Publications and source records attributed to Shengyu Li.

17 recordsLinked to original sources

arXiv2026

AudioScape-TTA: A Structured Soundscape Benchmark for Fine-Grained Text-to-Audio Evaluation

Text-to-audio (TTA) generation has recently achieved remarkable progress in synthesizing realistic audio from natural language descriptions. However, determining whether generated audio faithfully satisfies complex textual instructions remains challenging. Existing benchmarks mainly rely on global similarity metrics, providing limited insight into fine-grained semantic failures. To address this limitation, we introduce \textbf{AudioScape-TTA}, a structured and complexity-aware benchmark for fine-grained TTA evaluation. AudioScape-TTA represents realistic soundscapes through modality-aware semantic structures and characterizes generation complexity using event density and structural complexity. Based on these annotations, we propose a rubric-based audio-grounded evaluation framework that verifies event realization, acoustic attributes, and speech content through fine-grained semantic criteria. The benchmark contains 2,258 audio-text pairs with 25,707 binary QA rubrics, enabling scalable and interpretable analysis of TTA systems. Experiments on 13 representative open-source TTA models reveal persistent limitations in fine-grained attribute control, speech-content preservation, and compositional soundscape generation. Human validation further demonstrates that our rubric-based evaluation achieves stronger alignment with human semantic judgments than conventional global similarity metrics.

cs.SD↗

arXiv2025

PhyAVBench: A Challenging Audio Physics-Sensitivity Benchmark for Physically Grounded Text-to-Audio-Video Generation

Text-to-audio-video (T2AV) generation is central to applications such as filmmaking and world modeling. However, current models often fail to produce physically plausible sounds. Previous benchmarks primarily focus on audio-video temporal synchronization, while largely overlooking explicit evaluation of audio-physics grounding, thereby limiting the study of physically plausible audio-visual generation. To address this issue, we present PhyAVBench, the first benchmark that systematically evaluates the audio-physics grounding capabilities of T2AV, image-to-audio-video (I2AV), and video-to-audio (V2A) models. PhyAVBench offers PhyAV-Sound-11K, a new dataset of 25.5 hours of 11,605 audible videos collected from 184 participants to ensure diversity and avoid data leakage. It contains 337 paired-prompt groups with controlled physical variations that drive sound differences, each grounded with an average of 17 videos and spanning 6 audio-physics dimensions and 41 fine-grained test points. Each prompt pair is annotated with the physical factors underlying their acoustic differences. Importantly, PhyAVBench leverages paired text prompts to evaluate this capability. We term this evaluation paradigm the Audio-Physics Sensitivity Test (APST) and introduce a novel metric, the Contrastive Physical Response Score (CPRS), which quantifies the acoustic consistency between generated videos and their real-world counterparts. We conduct a comprehensive evaluation of 17 state-of-the-art models. Our results reveal that even leading commercial models struggle with fundamental audio-physical phenomena, exposing a critical gap beyond audio-visual synchronization and pointing to future research directions. We hope PhyAVBench will serve as a foundation for advancing physically grounded audio-visual generation. Prompts, ground-truth, and generated video samples are available at https://github.com/imxtx/PhyAVBench.

cs.SD↗

arXiv2025

Boosting Multi-modal Keyphrase Prediction with Dynamic Chain-of-Thought in Vision-Language Models

Multi-modal keyphrase prediction (MMKP) aims to advance beyond text-only methods by incorporating multiple modalities of input information to produce a set of conclusive phrases. Traditional multi-modal approaches have been proven to have significant limitations in handling the challenging absence and unseen scenarios. Additionally, we identify shortcomings in existing benchmarks that overestimate model capability due to significant overlap in training tests. In this work, we propose leveraging vision-language models (VLMs) for the MMKP task. Firstly, we use two widely-used strategies, e.g., zero-shot and supervised fine-tuning (SFT) to assess the lower bound performance of VLMs. Next, to improve the complex reasoning capabilities of VLMs, we adopt Fine-tune-CoT, which leverages high-quality CoT reasoning data generated by a teacher model to finetune smaller models. Finally, to address the "overthinking" phenomenon, we propose a dynamic CoT strategy which adaptively injects CoT data during training, allowing the model to flexibly leverage its reasoning capabilities during the inference stage. We evaluate the proposed strategies on various datasets and the experimental results demonstrate the effectiveness of the proposed approaches. The code is available at https://github.com/bytedance/DynamicCoT.

cs.CV↗

arXiv2025

MASt3R-Fusion: Integrating Feed-Forward Visual Model with IMU, GNSS for High-Functionality SLAM

Visual SLAM is a cornerstone technique in robotics, autonomous driving and extended reality (XR), yet classical systems often struggle with low-texture environments, scale ambiguity, and degraded performance under challenging visual conditions. Recent advancements in feed-forward neural network-based pointmap regression have demonstrated the potential to recover high-fidelity 3D scene geometry directly from images, leveraging learned spatial priors to overcome limitations of traditional multi-view geometry methods. However, the widely validated advantages of probabilistic multi-sensor information fusion are often discarded in these pipelines. In this work, we propose MASt3R-Fusion,a multi-sensor-assisted visual SLAM framework that tightly integrates feed-forward pointmap regression with complementary sensor information, including inertial measurements and GNSS data. The system introduces Sim(3)-based visualalignment constraints (in the Hessian form) into a universal metric-scale SE(3) factor graph for effective information fusion. A hierarchical factor graph design is developed, which allows both real-time sliding-window optimization and global optimization with aggressive loop closures, enabling real-time pose tracking, metric-scale structure perception and globally consistent mapping. We evaluate our approach on both public benchmarks and self-collected datasets, demonstrating substantial improvements in accuracy and robustness over existing visual-centered multi-sensor SLAM systems. The code will be released open-source to support reproducibility and further research (https://github.com/GREAT-WHU/MASt3R-Fusion).

cs.RO↗

arXiv2025

Uni-AIMS: AI-Powered Microscopy Image Analysis

This paper presents a systematic solution for the intelligent recognition and automatic analysis of microscopy images. We developed a data engine that generates high-quality annotated datasets through a combination of the collection of diverse microscopy images from experiments, synthetic data generation and a human-in-the-loop annotation process. To address the unique challenges of microscopy images, we propose a segmentation model capable of robustly detecting both small and large objects. The model effectively identifies and separates thousands of closely situated targets, even in cluttered visual environments. Furthermore, our solution supports the precise automatic recognition of image scale bars, an essential feature in quantitative microscopic analysis. Building upon these components, we have constructed a comprehensive intelligent analysis platform and validated its effectiveness and practicality in real-world applications. This study not only advances automatic recognition in microscopy imaging but also ensures scalability and generalizability across multiple application domains, offering a powerful tool for automated microscopic analysis in interdisciplinary research. A online application is made available for researchers to access and evaluate the proposed automated analysis service.

eess.IV↗

arXiv2025

S3MOT: Monocular 3D Object Tracking with Selective State Space Model

Accurate and reliable multi-object tracking (MOT) in 3D space is essential for advancing robotics and computer vision applications. However, it remains a significant challenge in monocular setups due to the difficulty of mining 3D spatiotemporal associations from 2D video streams. In this work, we present three innovative techniques to enhance the fusion and exploitation of heterogeneous cues for monocular 3D MOT: (1) we introduce the Hungarian State Space Model (HSSM), a novel data association mechanism that compresses contextual tracking cues across multiple paths, enabling efficient and comprehensive assignment decisions with linear complexity. HSSM features a global receptive field and dynamic weights, in contrast to traditional linear assignment algorithms that rely on hand-crafted association costs. (2) We propose Fully Convolutional One-stage Embedding (FCOE), which eliminates ROI pooling by directly using dense feature maps for contrastive learning, thus improving object re-identification accuracy under challenging conditions such as varying viewpoints and lighting. (3) We enhance 6-DoF pose estimation through VeloSSM, an encoder-decoder architecture that models temporal dependencies in velocity to capture motion dynamics, overcoming the limitations of frame-based 3D inference. Experiments on the KITTI public test benchmark demonstrate the effectiveness of our method, achieving a new state-of-the-art performance of 76.86~HOTA at 31~FPS. Our approach outperforms the previous best by significant margins of +2.63~HOTA and +3.62~AssA, showcasing its robustness and efficiency for monocular 3D MOT tasks. The code and models are available at https://github.com/bytepioneerX/s3mot.

cs.CV↗

arXiv2025

Prescribing hyperbolic bordered surfaces via combinatorial flows

The aim of this paper is to investigate the fractional combinatorial Calabi flow for hyperbolic bordered surfaces. By Lyapunov theory, it is proved that the flow exists for all time and converges exponentially to a conformal factor that generates a hyperbolic surface whose lengths of boundary components are prescribed positive numbers. Furthermore, a generalized combinatorial Yamabe flow is introduced in the same geometry setting, with the long time existence and convergence established. This result yields an algorithm for searching bordered surfaces, which may accelerate convergence speed.

math.CV↗

arXiv2025

Combinatorial Calabi flow for ideal circle pattern

We study the combinatorial Calabi flow for ideal circle patterns in both hyperbolic and Euclidean background geometry. We prove that the flow exists for all time and converges exponentially fast to an ideal circle pattern metric on surfaces with prescribed attainable curvatures. As a consequence, we provide an algorithm to find the desired ideal circle patterns.

math.DG↗

arXiv2024

SF-Loc: A Visual Mapping and Geo-Localization System based on Sparse Visual Structure Frames

For high-level geo-spatial applications and intelligent robotics, accurate global pose information is of crucial importance. Map-aided localization is a universal approach to overcome the limitations of global navigation satellite system (GNSS) in challenging environments. However, current solutions face challenges in terms of mapping flexibility, storage burden and re-localization performance. In this work, we present SF-Loc, a lightweight visual mapping and map-aided localization system, whose core idea is the map representation based on sparse frames with dense but compact depth, termed as visual structure frames. In the mapping phase, multi-sensor dense bundle adjustment (MS-DBA) is applied to construct geo-referenced visual structure frames. The local co-visbility is checked to keep the map sparsity and achieve incremental mapping. In the localization phase, coarse-to-fine vision-based localization is performed, in which multi-frame information and the map distribution are fully integrated. To be specific, the concept of spatially smoothed similarity (SSS) is proposed to overcome the place ambiguity, and pairwise frame matching is applied for efficient and robust pose estimation. Experimental results on the cross-season dataset verify the effectiveness of the system. In complex urban road scenarios, the map size is down to 3 MB per kilometer and stable decimeter-level re-localization can be achieved. The code will be made open-source soon (https://github.com/GREAT-WHU/SF-Loc).

cs.RO↗

arXiv2024

iKalibr-RGBD: Partially-Specialized Target-Free Visual-Inertial Spatiotemporal Calibration For RGBDs via Continuous-Time Velocity Estimation

Visual-inertial systems have been widely studied and applied in the last two decades (from the early 2000s to the present), mainly due to their low cost and power consumption, small footprint, and high availability. Such a trend simultaneously leads to a large amount of visual-inertial calibration methods being presented, as accurate spatiotemporal parameters between sensors are a prerequisite for visual-inertial fusion. In our previous work, i.e., iKalibr, a continuous-time-based visual-inertial calibration method was proposed as a part of one-shot multi-sensor resilient spatiotemporal calibration. While requiring no artificial target brings considerable convenience, computationally expensive pose estimation is demanded in initialization and batch optimization, limiting its availability. Fortunately, this could be vastly improved for the RGBDs with additional depth information, by employing mapping-free ego-velocity estimation instead of mapping-based pose estimation. In this paper, we present the continuous-time ego-velocity estimation-based RGBD-inertial spatiotemporal calibration, termed as iKalibr-RGBD, which is also targetless but computationally efficient. The general pipeline of iKalibr-RGBD is inherited from iKalibr, composed of a rigorous initialization procedure and several continuous-time batch optimizations. The implementation of iKalibr-RGBD is open-sourced at (https://github.com/Unsigned-Long/iKalibr) to benefit the research community.

cs.RO↗

arXiv2024

RIs-Calib: An Open-Source Spatiotemporal Calibrator for Multiple 3D Radars and IMUs Based on Continuous-Time Estimation

Aided inertial navigation system (INS), typically consisting of an inertial measurement unit (IMU) and an exteroceptive sensor, has been widely accepted as a feasible solution for navigation. Compared with vision-aided and LiDAR-aided INS, radar-aided INS could achieve better performance in adverse weather conditions since the radar utilizes low-frequency measuring signals with less attenuation effect in atmospheric gases and rain. For such a radar-aided INS, accurate spatiotemporal transformation is a fundamental prerequisite to achieving optimal information fusion. In this work, we present RIs-Calib: a spatiotemporal calibrator for multiple 3D radars and IMUs based on continuous-time estimation, which enables accurate spatiotemporal calibration and does not require any additional artificial infrastructure or prior knowledge. Our approach starts with a rigorous and robust procedure for state initialization, followed by batch optimizations, where all parameters can be refined to global optimal states steadily. We validate and evaluate RIs-Calib on both simulated and real-world experiments, and the results demonstrate that RIs-Calib is capable of accurate and consistent calibration. We open-source our implementations at (https://github.com/Unsigned-Long/RIs-Calib) to benefit the research community.

cs.RO↗

arXiv2024

iKalibr: Unified Targetless Spatiotemporal Calibration for Resilient Integrated Inertial Systems

The integrated inertial system, typically integrating an IMU and an exteroceptive sensor such as radar, LiDAR, and camera, has been widely accepted and applied in modern robotic applications for ego-motion estimation, motion control, or autonomous exploration. To improve system accuracy, robustness, and further usability, both multiple and various sensors are generally resiliently integrated, which benefits the system performance regarding failure tolerance, perception capability, and environment compatibility. For such systems, accurate and consistent spatiotemporal calibration is required to maintain a unique spatiotemporal framework for multi-sensor fusion. Considering most existing calibration methods (i) are generally oriented to specific integrated inertial systems, (ii) often only focus on spatial determination, (iii) usually require artificial targets, lacking convenience and usability, we propose iKalibr: a unified targetless spatiotemporal calibration framework for resilient integrated inertial systems, which overcomes the above issues, and enables both accurate and consistent calibration. Altogether four commonly employed sensors are supported in iKalibr currently, namely IMU, radar, LiDAR, and camera. The proposed method starts with a rigorous and efficient dynamic initialization, where all parameters in the estimator would be accurately recovered. Subsequently, several continuous-time batch optimizations are conducted to refine the initialized parameters toward better states. Sufficient real-world experiments were conducted to verify the feasibility and evaluate the calibration performance of iKalibr. The results demonstrate that iKalibr can achieve accurate resilient spatiotemporal calibration. We open-source our implementations at (https://github.com/Unsigned-Long/iKalibr) to benefit the research community.

cs.RO↗

arXiv2024

DBA-Fusion: Tightly Integrating Deep Dense Visual Bundle Adjustment with Multiple Sensors for Large-Scale Localization and Mapping

Visual simultaneous localization and mapping (VSLAM) has broad applications, with state-of-the-art methods leveraging deep neural networks for better robustness and applicability. However, there is a lack of research in fusing these learning-based methods with multi-sensor information, which could be indispensable to push related applications to large-scale and complex scenarios. In this paper, we tightly integrate the trainable deep dense bundle adjustment (DBA) with multi-sensor information through a factor graph. In the framework, recurrent optical flow and DBA are performed among sequential images. The Hessian information derived from DBA is fed into a generic factor graph for multi-sensor fusion, which employs a sliding window and supports probabilistic marginalization. A pipeline for visual-inertial integration is firstly developed, which provides the minimum ability of metric-scale localization and mapping. Furthermore, other sensors (e.g., global navigation satellite system) are integrated for driftless and geo-referencing functionality. Extensive tests are conducted on both public datasets and self-collected datasets. The results validate the superior localization performance of our approach, which enables real-time dense mapping in large-scale environments. The code has been made open-source (https://github.com/GREAT-WHU/DBA-Fusion).

cs.RO↗

arXiv2023

Ground-VIO: Monocular Visual-Inertial Odometry with Online Calibration of Camera-Ground Geometric Parameters

Monocular visual-inertial odometry (VIO) is a low-cost solution to provide high-accuracy, low-drifting pose estimation. However, it has been meeting challenges in vehicular scenarios due to limited dynamics and lack of stable features. In this paper, we propose Ground-VIO, which utilizes ground features and the specific camera-ground geometry to enhance monocular VIO performance in realistic road environments. In the method, the camera-ground geometry is modeled with vehicle-centered parameters and integrated into an optimization-based VIO framework. These parameters could be calibrated online and simultaneously improve the odometry accuracy by providing stable scale-awareness. Besides, a specially designed visual front-end is developed to stably extract and track ground features via the inverse perspective mapping (IPM) technique. Both simulation tests and real-world experiments are conducted to verify the effectiveness of the proposed method. The results show that our implementation could dramatically improve monocular VIO accuracy in vehicular scenarios, achieving comparable or even better performance than state-of-art stereo VIO solutions. The system could also be used for the auto-calibration of IPM which is widely used in vehicle perception. A toolkit for ground feature processing, together with the experimental datasets, would be made open-source (https://github.com/GREAT-WHU/gv_tools).

cs.RO↗

arXiv2022

Rigidity of bordered polyhedral surfaces

This paper investigates the rigidity of bordered polyhedral surfaces. Using the variational principle, we show that bordered polyhedral surfaces are determined by boundary value and discrete curvatures on the interior edges. As a corollary, we reprove the classical result that two Euclidean cyclic polygons (or hyperbolic cyclic polygons) are congruent if the lengths of their sides are equal.

math.GT↗

arXiv2022

Combinatorial Yamabe flow on hyperbolic bordered surfaces

This paper studies the combinatorial Yamabe flow on hyperbolic bordered surfaces. We show that the flow exists for all time and converges exponentially fast to conformal factor which produces a hyperbolic surface whose lengths of boundary components are equal to prescribed positive numbers. This provides an algorithm to such problems.

math.DG↗

arXiv2022

Dynamic Supervisor for Cross-dataset Object Detection

The application of cross-dataset training in object detection tasks is complicated because the inconsistency in the category range across datasets transforms fully supervised learning into semi-supervised learning. To address this problem, recent studies focus on the generation of high-quality missing annotations. In this study, we first point out that it is not enough to generate high-quality annotations using a single model, which only looks once for annotations. Through detailed experimental analyses, we further conclude that hard-label training is conducive to generating high-recall annotations, while soft-label training tends to obtain high-precision annotations. Inspired by the aspects mentioned above, we propose a dynamic supervisor framework that updates the annotations multiple times through multiple-updated submodels trained using hard and soft labels. In the final generated annotations, both recall and precision improve significantly through the integration of hard-label training with soft-label training. Extensive experiments conducted on various dataset combination settings support our analyses and demonstrate the superior performance of the proposed dynamic supervisor.

cs.CV↗