arXiv subjects

Zeyu Ma

Publications and source records attributed to Zeyu Ma.

At least 37 records · Page 2Linked to original sources

arXiv2025

ORCA: Object Recognition and Comprehension for Archiving Marine Species

Marine visual understanding is essential for monitoring and protecting marine ecosystems, enabling automatic and scalable biological surveys. However, progress is hindered by limited training data and the lack of a systematic task formulation that aligns domain-specific marine challenges with well-defined computer vision tasks, thereby limiting effective model application. To address this gap, we present ORCA, a multi-modal benchmark for marine research comprising 14,647 images from 478 species, with 42,217 bounding box annotations and 22,321 expert-verified instance captions. The dataset provides fine-grained visual and textual annotations that capture morphology-oriented attributes across diverse marine species. To catalyze methodological advances, we evaluate 18 state-of-the-art models on three tasks: object detection (closed-set and open-vocabulary), instance captioning, and visual grounding. Results highlight key challenges, including species diversity, morphological overlap, and specialized domain demands, underscoring the difficulty of marine understanding. ORCA thus establishes a comprehensive benchmark to advance research in marine domain. Project Page: http://orca.hkustvgd.com/.

cs.CV↗

arXiv2025

Temporally Smooth Mesh Extraction for Procedural Scenes with Long-Range Camera Trajectories using Spacetime Octrees

The procedural occupancy function is a flexible and compact representation for creating 3D scenes. For rasterization and other tasks, it is often necessary to extract a mesh that represents the shape. Unbounded scenes with long-range camera trajectories, such as flying through a forest, pose a unique challenge for mesh extraction. A single static mesh representing all the geometric detail necessary for the full camera path can be prohibitively large. Therefore, independent meshes can be extracted for different camera views, but this approach may lead to popping artifacts during transitions. We propose a temporally coherent method for extracting meshes suitable for long-range camera trajectories in unbounded scenes represented by an occupancy function. The key idea is to perform 4D mesh extraction using a new spacetime tree structure called a binary-octree. Experiments show that, compared to existing baseline methods, our method offers superior visual consistency at a comparable cost. The code and the supplementary video for this paper are available at https://github.com/princeton-vl/BinocMesher.

cs.GR↗

arXiv2025

Dendritic Resonate-and-Fire Neuron for Effective and Efficient Long Sequence Modeling

The explosive growth in sequence length has intensified the demand for effective and efficient long sequence modeling. Benefiting from intrinsic oscillatory membrane dynamics, Resonate-and-Fire (RF) neurons can efficiently extract frequency components from input signals and encode them into spatiotemporal spike trains, making them well-suited for long sequence modeling. However, RF neurons exhibit limited effective memory capacity and a trade-off between energy efficiency and training speed on complex temporal tasks. Inspired by the dendritic structure of biological neurons, we propose a Dendritic Resonate-and-Fire (D-RF) model, which explicitly incorporates a multi-dendritic and soma architecture. Each dendritic branch encodes specific frequency bands by utilizing the intrinsic oscillatory dynamics of RF neurons, thereby collectively achieving comprehensive frequency representation. Furthermore, we introduce an adaptive threshold mechanism into the soma structure that adjusts the threshold based on historical spiking activity, reducing redundant spikes while maintaining training efficiency in long sequence tasks. Extensive experiments demonstrate that our method maintains competitive accuracy while substantially ensuring sparse spikes without compromising computational efficiency during training. These results underscore its potential as an effective and efficient solution for long sequence modeling on edge platforms.

cs.LG↗

arXiv2025

TestAgent: Automatic Benchmarking and Exploratory Interaction for Evaluating LLMs in Vertical Domains

As Large Language Models (LLMs) are increasingly deployed in highly specialized vertical domains, the evaluation of their domain-specific performance becomes critical. However, existing evaluations for vertical domains typically rely on the labor-intensive construction of static single-turn datasets, which present two key limitations: (i) manual data construction is costly and must be repeated for each new domain, and (ii) static single-turn evaluations are misaligned with the dynamic multi-turn interactions in real-world applications, limiting the assessment of professionalism and stability. To address these, we propose TestAgent, a framework for automatic benchmarking and exploratory dynamic evaluation in vertical domains. TestAgent leverages retrieval-augmented generation to create domain-specific questions from user-provided knowledge sources, combined with a two-stage criteria generation process, thereby enabling scalable and automated benchmark creation. Furthermore, it introduces a reinforcement learning-guided multi-turn interaction strategy that adaptively determines question types based on real-time model responses, dynamically probing knowledge boundaries and stability. Extensive experiments across medical, legal, and governmental domains demonstrate that TestAgent enables efficient cross-domain benchmark generation and yields deeper insights into model behavior through dynamic exploratory evaluation. This work establishes a new paradigm for automated and in-depth evaluation of LLMs in vertical domains.

cs.AI↗

arXiv2025

Self-Aware Safety Augmentation: Leveraging Internal Semantic Understanding to Enhance Safety in Vision-Language Models

Large vision-language models (LVLMs) are vulnerable to harmful input compared to their language-only backbones. We investigated this vulnerability by exploring LVLMs internal dynamics, framing their inherent safety understanding in terms of three key capabilities. Specifically, we define these capabilities as safety perception, semantic understanding, and alignment for linguistic expression, and experimentally pinpointed their primary locations within the model architecture. The results indicate that safety perception often emerges before comprehensive semantic understanding, leading to the reduction in safety. Motivated by these findings, we propose \textbf{Self-Aware Safety Augmentation (SASA)}, a technique that projects informative semantic representations from intermediate layers onto earlier safety-oriented layers. This approach leverages the model's inherent semantic understanding to enhance safety recognition without fine-tuning. Then, we employ linear probing to articulate the model's internal semantic comprehension to detect the risk before the generation process. Extensive experiments on various datasets and tasks demonstrate that SASA significantly improves the safety of LVLMs, with minimal impact on the utility.

cs.AI↗

arXiv2025

CoCo: A Coupled Contrastive Framework for Unsupervised Domain Adaptive Graph Classification

Although graph neural networks (GNNs) have achieved impressive achievements in graph classification, they often need abundant task-specific labels, which could be extensively costly to acquire. A credible solution is to explore additional labeled graphs to enhance unsupervised learning on the target domain. However, how to apply GNNs to domain adaptation remains unsolved owing to the insufficient exploration of graph topology and the significant domain discrepancy. In this paper, we propose Coupled Contrastive Graph Representation Learning (CoCo), which extracts the topological information from coupled learning branches and reduces the domain discrepancy with coupled contrastive learning. CoCo contains a graph convolutional network branch and a hierarchical graph kernel network branch, which explore graph topology in implicit and explicit manners. Besides, we incorporate coupled branches into a holistic multi-view contrastive learning framework, which not only incorporates graph representations learned from complementary views for enhanced understanding, but also encourages the similarity between cross-domain example pairs with the same semantics for domain alignment. Extensive experiments on popular datasets show that our CoCo outperforms these competing baselines in different settings generally.

cs.LG↗

arXiv2025

Evaluating Robustness of Monocular Depth Estimation with Procedural Scene Perturbations

Recent years have witnessed substantial progress on monocular depth estimation, particularly as measured by the success of large models on standard benchmarks. However, performance on standard benchmarks does not offer a complete assessment, because most evaluate accuracy but not robustness. In this work, we introduce PDE (Procedural Depth Evaluation), a new benchmark which enables systematic robustness evaluation. PDE uses procedural generation to create 3D scenes that test robustness to various controlled perturbations, including object, camera, material and lighting changes. Our analysis yields interesting findings on what perturbations are challenging for state-of-the-art depth models, which we hope will inform further research. Code and data are available at https://github.com/princeton-vl/proc-depth-eval.

cs.CV↗

arXiv2025

OMNI-DC: Highly Robust Depth Completion with Multiresolution Depth Integration

Depth completion (DC) aims to predict a dense depth map from an RGB image and a sparse depth map. Existing DC methods generalize poorly to new datasets or unseen sparse depth patterns, limiting their real-world applications. We propose OMNI-DC, a highly robust DC model that generalizes well zero-shot to various datasets. The key design is a novel Multi-resolution Depth Integrator, allowing our model to deal with very sparse depth inputs. We also introduce a novel Laplacian loss to model the ambiguity in the training process. Moreover, we train OMNI-DC on a mixture of high-quality datasets with a scale normalization technique and synthetic depth patterns. Extensive experiments on 7 datasets show consistent improvements over baselines, reducing errors by as much as 43%. Codes and checkpoints are available at https://github.com/princeton-vl/OMNI-DC.

cs.CV↗

arXiv2024

SVFit: Parameter-Efficient Fine-Tuning of Large Pre-Trained Models Using Singular Values

Large pre-trained models (LPMs) have demonstrated exceptional performance in diverse natural language processing and computer vision tasks. However, fully fine-tuning these models poses substantial memory challenges, particularly in resource-constrained environments. Parameter-efficient fine-tuning (PEFT) methods, such as LoRA, mitigate this issue by adjusting only a small subset of parameters. Nevertheless, these methods typically employ random initialization for low-rank matrices, which can lead to inefficiencies in gradient descent and diminished generalizability due to suboptimal starting points. To address these limitations, we propose SVFit, a novel PEFT approach that leverages singular value decomposition (SVD) to initialize low-rank matrices using critical singular values as trainable parameters. Specifically, SVFit performs SVD on the pre-trained weight matrix to obtain the best rank-r approximation matrix, emphasizing the most critical singular values that capture over 99% of the matrix's information. These top-r singular values are then used as trainable parameters to scale the fundamental subspaces of the matrix, facilitating rapid domain adaptation. Extensive experiments across various pre-trained models in natural language understanding, text-to-image generation, and image classification tasks reveal that SVFit outperforms LoRA while requiring 16 times fewer trainable parameters.

cs.LG↗

arXiv2024

Through-the-Wall Radar Human Activity Micro-Doppler Signature Representation Method Based on Joint Boulic-Sinusoidal Pendulum Model

With the help of micro-Doppler signature, ultra-wideband (UWB) through-the-wall radar (TWR) enables the reconstruction of range and velocity information of limb nodes to accurately identify indoor human activities. However, existing methods are usually trained and validated directly using range-time maps (RTM) and Doppler-time maps (DTM), which have high feature redundancy and poor generalization ability. In order to solve this problem, this paper proposes a human activity micro-Doppler signature representation method based on joint Boulic-sinusoidal pendulum motion model. In detail, this paper presents a simplified joint Boulic-sinusoidal pendulum human motion model by taking head, torso, both hands and feet into consideration improved from Boulic-Thalmann kinematic model. The paper also calculates the minimum number of key points needed to describe the Doppler and micro-Doppler information sufficiently. Both numerical simulations and experiments are conducted to verify the effectiveness. The results demonstrate that the proposed number of key points of micro-Doppler signature can precisely represent the indoor human limb node motion characteristics, and substantially improve the generalization capability of the existing methods for different testers.

eess.SP↗

arXiv2024

Anisotropic magnetic interactions in a candidate Kitaev spin liquid close to a metal-insulator transition

In the Kitaev honeycomb model, spins coupled by strongly-frustrated anisotropic interactions do not order at low temperature but instead form a quantum spin liquid with spin fractionalization into Majorana fermions and static fluxes. The realization of such a model in crystalline materials could lead to major breakthroughs in understanding entangled quantum states, however achieving this in practice is a very challenging task. The recently synthesized honeycomb material RuI$_3$ shows no long-range magnetic order down to the lowest probed temperatures and has been theoretically proposed as a quantum spin liquid candidate material on the verge of an insulator to metal transition. Here we report a comprehensive study of the magnetic anisotropy in un-twinned single crystals via torque magnetometry and detect clear signatures of strongly anisotropic and frustrated magnetic interactions. We attribute the development of sawtooth and six-fold torque signal to strongly anisotropic, bond-dependent magnetic interactions by comparing to theoretical calculations. As a function of magnetic field strength at low temperatures, torque shows an unusual non-parabolic dependence suggestive of a proximity to a field-induced transition. Thus, RuI$_3$, without signatures of long-range magnetic order, displays key hallmarks of an exciting new candidate for extended Kitaev magnetism with enhanced quantum fluctuations.

cond-mat.str-el↗

arXiv2024

Anisotropy of the zigzag order in the Kitaev honeycomb magnet $α$-RuBr$_3$

Kitaev materials often order magnetically at low temperatures due to the presence of non-Kitaev interactions. Torque magnetometry is a very sensitive technique for probing the magnetic anisotropy, which is critical in understanding the magnetic ground state. In this work, we report detailed single-crystal torque measurements in the proposed Kitaev candidate honeycomb magnet $α$-RuBr$_3$, which displays zigzag order below 34 K. Based on angular-dependent torque studies in magnetic fields up to 16 T rotated in the plane normal to the honeycomb layers, we find an easy-plane anisotropy with a temperature dependence of the torque amplitude following closely the behaviour of the powder magnetic susceptibility. The torque for field rotated in the honeycomb plane has a clear six-fold periodicity with a saw-tooth shape, reflecting the three-fold symmetry of the crystal structure and stabilization of different zigzag domains depending on the field orientation, with a torque amplitude that follows an order parameter form inside the zigzag phase. By comparing experimental data with theoretical calculations we identify the relevant anisotropic interactions and the role of the competition between different zigzag domains in this candidate Kitaev material.

cond-mat.str-el↗

arXiv2024

Infinigen Indoors: Photorealistic Indoor Scenes using Procedural Generation

We introduce Infinigen Indoors, a Blender-based procedural generator of photorealistic indoor scenes. It builds upon the existing Infinigen system, which focuses on natural scenes, but expands its coverage to indoor scenes by introducing a diverse library of procedural indoor assets, including furniture, architecture elements, appliances, and other day-to-day objects. It also introduces a constraint-based arrangement system, which consists of a domain-specific language for expressing diverse constraints on scene composition, and a solver that generates scene compositions that maximally satisfy the constraints. We provide an export tool that allows the generated 3D objects and scenes to be directly used for training embodied agents in real-time simulators such as Omniverse and Unreal. Infinigen Indoors is open-sourced under the BSD license. Please visit https://infinigen.org for code and videos.

cs.CV↗

arXiv2024

1/3 and other magnetization plateaus in a quasi-one-dimensional Ising magnet $\mathbf{TbTi_3Bi_4}$ with zigzag spin chain

We report the magnetic properties of newly synthesized, single crystals of $\mathrm{TbTi_3Bi_4}$ whose crystal structure is highlighted by the stacking of terbium-based zigzag chains and titanium-based kagome lattices. This compound demonstrates extreme easy-axis magnetic anisotropy due to the crystalline-electric-field effect which aligns the $\mathrm{Tb^{3+}}$ moments along the zigzag chain direction. As the result of the strong single-ion anisotropy and multiple magnetic interactions, $\mathrm{TbTi_3Bi_4}$ behaves as a quasi-one-dimensional Ising magnet with a remarkable antiferromagnetic ordering at $T_\mathrm{N}$ = 20.4 K. When a magnetic field is applied along the direction of the zigzag chain, multiple meta-magnetic transitions occur between 1/3 and other magnetization plateaus. We have created a field-temperature phase diagram and mapped out the complex magnetic structures resulting from frustration.

cond-mat.str-el↗

arXiv2023

View-Dependent Octree-based Mesh Extraction in Unbounded Scenes for Procedural Synthetic Data

Procedural synthetic data generation has received increasing attention in computer vision. Procedural signed distance functions (SDFs) are a powerful tool for modeling large-scale detailed scenes, but existing mesh extraction methods have artifacts or performance profiles that limit their use for synthetic data. We propose OcMesher, a mesh extraction algorithm that efficiently handles high-detail unbounded scenes with perfect view-consistency, with easy export to downstream real-time engines. The main novelty of our solution is an algorithm to construct an octree based on a given SDF and multiple camera views. We performed extensive experiments, and show our solution produces better synthetic data for training and evaluation of computer vision models.

cs.CV↗

arXiv2023

Infinite Photorealistic Worlds using Procedural Generation

We introduce Infinigen, a procedural generator of photorealistic 3D scenes of the natural world. Infinigen is entirely procedural: every asset, from shape to texture, is generated from scratch via randomized mathematical rules, using no external source and allowing infinite variation and composition. Infinigen offers broad coverage of objects and scenes in the natural world including plants, animals, terrains, and natural phenomena such as fire, cloud, rain, and snow. Infinigen can be used to generate unlimited, diverse training data for a wide range of computer vision tasks including object detection, semantic segmentation, optical flow, and 3D reconstruction. We expect Infinigen to be a useful resource for computer vision research and beyond. Please visit https://infinigen.org for videos, code and pre-generated data.

cs.CV↗

arXiv2022

Semantic Enhanced Knowledge Graph for Large-Scale Zero-Shot Learning

Zero-Shot Learning has been a highlighted research topic in both vision and language areas. Recently, most existing methods adopt structured knowledge information to model explicit correlations among categories and use deep graph convolutional network to propagate information between different categories. However, it is difficult to add new categories to existing structured knowledge graph, and deep graph convolutional network suffers from over-smoothing problem. In this paper, we provide a new semantic enhanced knowledge graph that contains both expert knowledge and categories semantic correlation. Our semantic enhanced knowledge graph can further enhance the correlations among categories and make it easy to absorb new categories. To propagate information on the knowledge graph, we propose a novel Residual Graph Convolutional Network (ResGCN), which can effectively alleviate the problem of over-smoothing. Experiments conducted on the widely used large-scale ImageNet-21K dataset and AWA2 dataset show the effectiveness of our method, and establish a new state-of-the-art on zero-shot learning. Moreover, our results on the large-scale ImageNet-21K with various feature extraction networks show that our method has better generalization and robustness.

cs.CV↗

arXiv2022

Learning Modal-Invariant and Temporal-Memory for Video-based Visible-Infrared Person Re-Identification

Thanks for the cross-modal retrieval techniques, visible-infrared (RGB-IR) person re-identification (Re-ID) is achieved by projecting them into a common space, allowing person Re-ID in 24-hour surveillance systems. However, with respect to the probe-to-gallery, almost all existing RGB-IR based cross-modal person Re-ID methods focus on image-to-image matching, while the video-to-video matching which contains much richer spatial- and temporal-information remains under-explored. In this paper, we primarily study the video-based cross-modal person Re-ID method. To achieve this task, a video-based RGB-IR dataset is constructed, in which 927 valid identities with 463,259 frames and 21,863 tracklets captured by 12 RGB/IR cameras are collected. Based on our constructed dataset, we prove that with the increase of frames in a tracklet, the performance does meet more enhancement, demonstrating the significance of video-to-video matching in RGB-IR person Re-ID. Additionally, a novel method is further proposed, which not only projects two modalities to a modal-invariant subspace, but also extracts the temporal-memory for motion-invariant. Thanks to these two strategies, much better results are achieved on our video-based cross-modal person Re-ID. The code and dataset are released at: https://github.com/VCMproject233/MITML.

cs.CV↗