arXiv subjects

Ziyang Chen

Publications and source records attributed to Ziyang Chen.

At least 91 records · Page 5Linked to original sources

arXiv2025

Images that Sound: Composing Images and Sounds on a Single Canvas

Spectrograms are 2D representations of sound that look very different from the images found in our visual world. And natural images, when played as spectrograms, make unnatural sounds. In this paper, we show that it is possible to synthesize spectrograms that simultaneously look like natural images and sound like natural audio. We call these visual spectrograms images that sound. Our approach is simple and zero-shot, and it leverages pre-trained text-to-image and text-to-spectrogram diffusion models that operate in a shared latent space. During the reverse process, we denoise noisy latents with both the audio and image diffusion models in parallel, resulting in a sample that is likely under both models. Through quantitative evaluations and perceptual studies, we find that our method successfully generates spectrograms that align with a desired audio prompt while also taking the visual appearance of a desired image prompt. Please see our project page for video results: https://ificl.github.io/images-that-sound/

cs.CV↗

arXiv2025

Test Time Training for 4D Medical Image Interpolation

4D medical image interpolation is essential for improving temporal resolution and diagnostic precision in clinical applications. Previous works ignore the problem of distribution shifts, resulting in poor generalization under different distribution. A natural solution would be to adapt the model to a new test distribution, but this cannot be done if the test input comes without a ground truth label. In this paper, we propose a novel test time training framework which uses self-supervision to adapt the model to a new distribution without requiring any labels. Indeed, before performing frame interpolation on each test video, the model is trained on the same instance using a self-supervised task, such as rotation prediction or image reconstruction. We conduct experiments on two publicly available 4D medical image interpolation datasets, Cardiac and 4D-Lung. The experimental results show that the proposed method achieves significant performance across various evaluation metrics on both datasets. It achieves higher peak signal-to-noise ratio values, 33.73dB on Cardiac and 34.02dB on 4D-Lung. Our method not only advances 4D medical image interpolation but also provides a template for domain adaptation in other fields such as image segmentation and image registration.

eess.IV↗

arXiv2025

GPS as a Control Signal for Image Generation

We show that the GPS tags contained in photo metadata provide a useful control signal for image generation. We train GPS-to-image models and use them for tasks that require a fine-grained understanding of how images vary within a city. In particular, we train a diffusion model to generate images conditioned on both GPS and text. The learned model generates images that capture the distinctive appearance of different neighborhoods, parks, and landmarks. We also extract 3D models from 2D GPS-to-image models through score distillation sampling, using GPS conditioning to constrain the appearance of the reconstruction from each viewpoint. Our evaluations suggest that our GPS-conditioned models successfully learn to generate images that vary based on location, and that GPS conditioning improves estimated 3D structure.

cs.CV↗

arXiv2025

MaxMin-L2-SVC-NCH: A Novel Approach for Support Vector Classifier Training and Parameter Selection

The selection of Gaussian kernel parameters plays an important role in the applications of support vector classification (SVC). A commonly used method is the k-fold cross validation with grid search (CV), which is extremely time-consuming because it needs to train a large number of SVC models. In this paper, a new approach is proposed to train SVC and optimize the selection of Gaussian kernel parameters. We first formulate the training and the parameter selection of SVC as a minimax optimization problem named as MaxMin-L2-SVC-NCH, in which the minimization problem is an optimization problem of finding the closest points between two normal convex hulls (L2-SVC-NCH) while the maximization problem is an optimization problem of finding the optimal Gaussian kernel parameters. A lower time complexity can be expected in MaxMin-L2-SVC-NCH because CV is not needed. We then propose a projected gradient algorithm (PGA) for the training of L2-SVC-NCH. It is revealed that the famous sequential minimal optimization (SMO) algorithm is a special case of the PGA. Thus, the PGA can provide more flexibility than the SMO. Furthermore, the solution of the maximization problem is done by a gradient ascent algorithm with dynamic learning rate. The comparative experiments between MaxMin-L2-SVC-NCH and the previous best approaches on public datasets show that MaxMin-L2-SVC-NCH greatly reduces the number of models to be trained while maintaining competitive test accuracy. These findings indicate that MaxMin-L2-SVC-NCH is a better choice for SVC tasks.

cs.LG↗

arXiv2025

Touchstone Benchmark: Are We on the Right Way for Evaluating AI Algorithms for Medical Segmentation?

How can we test AI performance? This question seems trivial, but it isn't. Standard benchmarks often have problems such as in-distribution and small-size test sets, oversimplified metrics, unfair comparisons, and short-term outcome pressure. As a consequence, good performance on standard benchmarks does not guarantee success in real-world scenarios. To address these problems, we present Touchstone, a large-scale collaborative segmentation benchmark of 9 types of abdominal organs. This benchmark is based on 5,195 training CT scans from 76 hospitals around the world and 5,903 testing CT scans from 11 additional hospitals. This diverse test set enhances the statistical significance of benchmark results and rigorously evaluates AI algorithms across various out-of-distribution scenarios. We invited 14 inventors of 19 AI algorithms to train their algorithms, while our team, as a third party, independently evaluated these algorithms on three test sets. In addition, we also evaluated pre-existing AI frameworks--which, differing from algorithms, are more flexible and can support different algorithms--including MONAI from NVIDIA, nnU-Net from DKFZ, and numerous other open-source frameworks. We are committed to expanding this benchmark to encourage more innovation of AI algorithms for the medical domain.

cs.CV↗

arXiv2024

Gradient Alignment Improves Test-Time Adaptation for Medical Image Segmentation

Although recent years have witnessed significant advancements in medical image segmentation, the pervasive issue of domain shift among medical images from diverse centres hinders the effective deployment of pre-trained models. Many Test-time Adaptation (TTA) methods have been proposed to address this issue by fine-tuning pre-trained models with test data during inference. These methods, however, often suffer from less-satisfactory optimization due to suboptimal optimization direction (dictated by the gradient) and fixed step-size (predicated on the learning rate). In this paper, we propose the Gradient alignment-based Test-time adaptation (GraTa) method to improve both the gradient direction and learning rate in the optimization procedure. Unlike conventional TTA methods, which primarily optimize the pseudo gradient derived from a self-supervised objective, our method incorporates an auxiliary gradient with the pseudo one to facilitate gradient alignment. Such gradient alignment enables the model to excavate the similarities between different gradients and correct the gradient direction to approximate the empirical gradient related to the current segmentation task. Additionally, we design a dynamic learning rate based on the cosine similarity between the pseudo and auxiliary gradients, thereby empowering the adaptive fine-tuning of pre-trained models on diverse test data. Extensive experiments establish the effectiveness of the proposed gradient alignment and dynamic learning rate and substantiate the superiority of our GraTa method over other state-of-the-art TTA methods on a benchmark medical image segmentation task. The code and weights of pre-trained source models are available at https://github.com/Chen-Ziyang/GraTa.

cs.CV↗

arXiv2024

Inline-Amplification-Free Time Transfer Utilizing Waveform-Resolved Single-Photon Detection

High-precision time transfer over a long haul of fiber plays a significant role in many fields. The core method, namely cascading relay nodes for the compensation of signal attenuation and dispersion, is however insufficient to deal with crucial point-to-point transfer scenarios, such as harsh environments with extremely deficient infrastructure and emergency conditions. In long-distance signal transmission without any inline amplifiers, the high loss of the optical fiber link becomes the primary limiting factor, and direct use of traditional photodetectors at the receiving end will bring about a significant drop in the stability of detected signals. Here we propose a waveform-resolved single photon detection technique and experimentally perform tomography on the weak transferred signal with an average photon number of just 0.617 per pulse. By adopting this technique, we achieve the time deviation of 95.68 ps and 192.58 ps at 200 km and 300 km respectively at an averaging time of 1 s, overcoming the technical lower bound induced by traditional photodetectors. This work lays the foundation for through-type time transfer with high precision in those significant inline-amplification-free scenarios.

physics.optics↗

arXiv2024

Day-Night Adaptation: An Innovative Source-free Adaptation Framework for Medical Image Segmentation

Distribution shifts widely exist in medical images acquired from different medical centres, hindering the deployment of semantic segmentation models trained on one centre (source domain) to another (target domain). While unsupervised domain adaptation has shown significant promise in mitigating these shifts, it poses privacy risks due to sharing data between centres. To facilitate adaptation while preserving data privacy, source-free domain adaptation (SFDA) and test-time adaptation (TTA) have emerged as effective paradigms, relying solely on target domain data. However, SFDA requires a pre-collected target domain dataset before deployment. TTA insufficiently exploit the potential value of test data, as it processes the test data only once. Considering that most medical centres operate during the day and remain inactive at night in clinical practice, we propose a novel adaptation framework called Day-Night Adaptation (DyNA) with above insights, which performs adaptation through day-night cycles without requiring access to source data. During the day, a low-frequency prompt is trained to adapt the frozen model to each test sample. We construct a memory bank for prompt initialization and develop a warm-up mechanism to enhance prompt training. During the night, we reuse test data collected from the day and introduce a global student model to bridge the knowledge between teacher and student models, facilitating model fine-tuning while ensuring training stability. Extensive experiments demonstrate that our DyNA outperforms existing TTA and SFDA methods on two benchmark medical image segmentation tasks. Code will be available after the paper is published.

cs.CV↗

arXiv2024

CoSAM: Self-Correcting SAM for Domain Generalization in 2D Medical Image Segmentation

Medical images often exhibit distribution shifts due to variations in imaging protocols and scanners across different medical centers. Domain Generalization (DG) methods aim to train models on source domains that can generalize to unseen target domains. Recently, the segment anything model (SAM) has demonstrated strong generalization capabilities due to its prompt-based design, and has gained significant attention in image segmentation tasks. Existing SAM-based approaches attempt to address the need for manual prompts by introducing prompt generators that automatically generate these prompts. However, we argue that auto-generated prompts may not be sufficiently accurate under distribution shifts, potentially leading to incorrect predictions that still require manual verification and correction by clinicians. To address this challenge, we propose a method for 2D medical image segmentation called Self-Correcting SAM (CoSAM). Our approach begins by generating coarse masks using SAM in a prompt-free manner, providing prior prompts for the subsequent stages, and eliminating the need for prompt generators. To automatically refine these coarse masks, we introduce a generalized error decoder that simulates the correction process typically performed by clinicians. Furthermore, we generate diverse prompts as feedback based on the corrected masks, which are used to iteratively refine the predictions within a self-correcting loop, enhancing the generalization performance of our model. Extensive experiments on two medical image segmentation benchmarks across multiple scenarios demonstrate the superiority of CoSAM over state-of-the-art SAM-based methods.

cs.CV↗

arXiv2024

An Adaptive Framework for Generating Systematic Explanatory Answer in Online Q&A Platforms

Question Answering (QA) systems face challenges in handling complex questions that require multi-domain knowledge synthesis. The naive RAG models, although effective in information retrieval, struggle with complex questions that require comprehensive and in-depth answers. The pioneering task is defined as explanatory answer generation, which entails handling identified challenges such as the requirement for comprehensive information and logical coherence within the generated context. To address these issues, we refer to systematic thinking theory and propose SynthRAG, an innovative framework designed to enhance QA performance. SynthRAG improves on conventional models by employing adaptive outlines for dynamic content structuring, generating systematic information to ensure detailed coverage, and producing customized answers tailored to specific user inquiries. This structured approach guarantees logical coherence and thorough integration of information, yielding responses that are both insightful and methodically organized. Empirical evaluations underscore SynthRAG's effectiveness, demonstrating its superiority in handling complex questions, overcoming the limitations of naive RAG models, and significantly improving answer quality and depth. Furthermore, an online deployment on the Zhihu platform revealed that SynthRAG's answers achieved notable user engagement, with each response averaging 5.73 upvotes and surpassing the performance of 79.8% of human contributors, highlighting the practical relevance and impact of the proposed framework. Our code is available at https://github.com/czy1999/SynthRAG .

cs.CL↗

arXiv2024

MedUniSeg: 2D and 3D Medical Image Segmentation via a Prompt-driven Universal Model

Universal segmentation models offer significant potential in addressing a wide range of tasks by effectively leveraging discrete annotations. As the scope of tasks and modalities expands, it becomes increasingly important to generate and strategically position task- and modal-specific priors within the universal model. However, existing universal models often overlook the correlations between different priors, and the optimal placement and frequency of these priors remain underexplored. In this paper, we introduce MedUniSeg, a prompt-driven universal segmentation model designed for 2D and 3D multi-task segmentation across diverse modalities and domains. MedUniSeg employs multiple modal-specific prompts alongside a universal task prompt to accurately characterize the modalities and tasks. To generate the related priors, we propose the modal map (MMap) and the fusion and selection (FUSE) modules, which transform modal and task prompts into corresponding priors. These modal and task priors are systematically introduced at the start and end of the encoding process. We evaluate MedUniSeg on a comprehensive multi-modal upstream dataset consisting of 17 sub-datasets. The results demonstrate that MedUniSeg achieves superior multi-task segmentation performance, attaining a 1.2% improvement in the mean Dice score across the 17 upstream tasks compared to nnUNet baselines, while using less than 1/10 of the parameters. For tasks that underperform during the initial multi-task joint training, we freeze MedUniSeg and introduce new modules to re-learn these tasks. This approach yields an enhanced version, MedUniSeg*, which consistently outperforms MedUniSeg across all tasks. Moreover, MedUniSeg surpasses advanced self-supervised and supervised pre-trained models on six downstream tasks, establishing itself as a high-quality, highly generalizable pre-trained segmentation model.

cs.CV↗

arXiv2024

SpecTrack: Learned Multi-Rotation Tracking via Speckle Imaging

Precision pose detection is increasingly demanded in fields such as personal fabrication, Virtual Reality (VR), and robotics due to its critical role in ensuring accurate positioning information. However, conventional vision-based systems used in these systems often struggle with achieving high precision and accuracy, particularly when dealing with complex environments or fast-moving objects. To address these limitations, we investigate Laser Speckle Imaging (LSI), an emerging optical tracking method that offers promising potential for improving pose estimation accuracy. Specifically, our proposed LSI-Based Tracking (SpecTrack) leverages the captures from a lensless camera and a retro-reflector marker with a coded aperture to achieve multi-axis rotational pose estimation with high precision. Our extensive trials using our in-house built testbed have shown that SpecTrack achieves an accuracy of 0.31° (std=0.43°), significantly outperforming state-of-the-art approaches and improving accuracy up to 200%.

cs.ET↗

arXiv2024

Efficient k-step Weighted Reachability Query Processing Algorithms

Given a data graph G, a source vertex u and a target vertex v of a reachability query, the reachability query is used to answer whether there exists a path from u to v in G. Reachability query processing is one of the fundamental operations in graph data management, which is widely used in biological networks, communication networks, and social networks to assist data analysis. The data graphs in practical applications usually contain information such as quantization weights associated with the structural relationships, in addition to the structural relationships between vertices. Thus, in addition to the traditional reachability relationships, users may want to further understand whether such reachability relationships satisfy specific constraints. In this paper, we study the problem of efficiently processing k -step reachability queries with weighted constraints in weighted graphs. The k -step weighted reachability query questions are used to answer the question of whether there exists a path from a source vertex u to a goal vertex v in a given weighted graph. If it exists, the path needs to satisfy 1) all edges in the path satisfy the given weight constraints, and 2) the length of the path does not exceed the given distance threshold k. To address the problem, firstly, WKRI index supporting k -step weighted reachability query processing and index construction methods based on efficient pruning strategies are proposed. Secondly, the idea of constructing index based on part of the vertexs is proposed to reduce the size of the index. We design and implement two optimized indexes GWKRI and LWKRI based on the vertex coverage set. Finally, experiments are conducted on several real datasets. The experimental results verify the efficiency of the method proposed in this paper in answering k -step weighted reachability queries.

cs.DB↗

arXiv2024

Using Prompts to Guide Large Language Models in Imitating a Real Person's Language Style

Large language models (LLMs), such as GPT series and Llama series have demonstrated strong capabilities in natural language processing, contextual understanding, and text generation. In recent years, researchers are trying to enhance the abilities of LLMs in performing various tasks, and numerous studies have proved that well-designed prompts can significantly improve the performance of LLMs on these tasks. This study compares the language style imitation ability of three different large language models under the guidance of the same zero-shot prompt. It also involves comparing the imitation ability of the same large language model when guided by three different prompts individually. Additionally, by applying a Tree-of-Thoughts (ToT) Prompting method to Llama 3, a conversational AI with the language style of a real person was created. In this study, three evaluation methods were used to evaluate LLMs and prompts. The results show that Llama 3 performs best at imitating language styles, and that the ToT prompting method is the most effective to guide it in imitating language styles. Using a ToT framework, Llama 3 was guided to interact with users in the language style of a specific individual without altering its core parameters, thereby creating a text-based conversational AI that reflects the language style of the individual.

cs.CL↗

arXiv2024

Deep Learning Enhanced Quantum Holography with Undetected Photons

Holography is an essential technique of generating three-dimensional images. Recently, quantum holography with undetected photons (QHUP) has emerged as a groundbreaking method capable of capturing complex amplitude images. Despite its potential, the practical application of QHUP has been limited by susceptibility to phase disturbances, low interference visibility, and limited spatial resolution. Deep learning, recognized for its ability in processing complex data, holds significant promise in addressing these challenges. In this report, we present an ample advancement in QHUP achieved by harnessing the power of deep learning to extract images from single-shot holograms, resulting in vastly reduced noise and distortion, alongside a notable enhancement in spatial resolution. The proposed and demonstrated deep learning QHUP (DL-QHUP) methodology offers a transformative solution by delivering high-speed imaging, improved spatial resolution, and superior noise resilience, making it suitable for diverse applications across an array of research fields stretching from biomedical imaging to remote sensing. DL-QHUP signifies a crucial leap forward in the realm of holography, demonstrating its immense potential to revolutionize imaging capabilities and pave the way for advancements in various scientific disciplines. The integration of DL-QHUP promises to unlock new possibilities in imaging applications, transcending existing limitations and offering unparalleled performance in challenging environments.

physics.optics↗

arXiv2024

On the capability of high redshift kSZ measurement with galaxy surveys

The kSZ effect has been detected at z<1 using various techniques and data sets. The ongoing and upcoming spectroscopic galaxy surveys such as DESI and PFS will push the detection beyond z = 1, and therefore map the baryon distribution at high redshifts. Such detection can be achieved by both the kSZ stacking and tomography methods. While the two methods are theoretically equivalent, they differ significantly in the probed physics and scales, and required data sets. Taking the combination of PFS and ACT as an example, we build mocks of kSZ and galaxies, quantify the kSZ detection S/N, and compare between the two methods. We segment the PFS galaxies into three redshift bins: 0.6 < z < 1.0, 1.0 < z < 1.6, and 1.6 < z < 2.4. For tomography method, our analysis reveals that the two higher redshift bins exhibit higher S/N, with values of 32 and 28, respectively, compared to the first redshift bin (S/N = 8). This is attributed to not only the increasing of electron density with redshifts, but also the larger survey volume and the reduced non-linearity, facilitating velocity reconstruction at higher redshifts. Therefore, the capability of the PFS survey to measure high redshift kSZ effect stands as a substantial advantage over other spectroscopic surveys at lower redshift. The S/N of kSZ stacking largely depends on the number of clusters/groups available from another photometric survey. But in general, its S/N is lower than that of kSZ tomography. Incorporating next-generation CMB surveys like CMB-S4, characterized by significantly reduced instrument noise and improved angular resolution, is expected to enhance tomographic detection by a factor of ten and stacking detection by five. This future high S/N detection holds the promise of not only providing precise constraints on the overall baryon abundance but also initiating a new insight into baryon distribution.

astro-ph.CO↗

arXiv2024

Dual-comb-enhanced microwave clock synchronization over commercial fiber

The large-scale clock network is the key ingredient to obtain high precision in many scenarios, from fundamental research to cutting-edge applications. The advantage of the time synchronization among microwave clocks is their cost, size, and accessibility. Here, we demonstrate a femtosecond-level time synchronization of microwave clocks through a commercial link of 205.86 km via dual-comb-enhanced optical two-way time transfer, which achieves a 6.23-fs residual time deviation between synchronized timescales at 1 s and an instability below 6E-18 at 10,000 s. Further, the high-precision time synchronization of microwave clocks significantly enhances the probe ability of subtle reciprocity changes of fiber to the sub-picosecond level. This work provides a path toward secure fiber time-frequency networks to support future microwave-clock-based precise timing and sensing systems.

physics.optics↗

arXiv2024

Gain-Bandwidth-Product-Induced Technical Bound in Time Transfer System without Inline Amplifiers

Time transfer plays a dispensable role in many fields including navigation and positioning, geodesy, and fundamental tests. However, in certain scenarios where effective relay node deployment is often not feasible, such as harsh environments with extremely poor infrastructure and emergency conditions, effective time transfer without inline amplifiers becomes crucial. In this situation, the maximum transmission distance is limited by the receiver's measurement capability, particularly its ability to amplify the signal. Here we propose a theoretical model, giving a technical lower bound of the detected signal stability at different transmission distances, induced by limited gain-bandwidth products. The results under common gain-bandwidth products show that while for shorter transmission distances, stability is mainly limited by the background noise of the time interval counter, for longer distances reaching the scale of 300 kilometers, the technical lower bound is below the level of 10 nanoseconds without any inline amplification devices. Therefore, the given technical bound offers guidance on managing the balance between distance and stability, together with the optimization of the receiver in long-distance time transfer without inline amplification.

physics.ins-det↗