arXiv subjects

Chenghao Xu

Publications and source records attributed to Chenghao Xu.

34 records · Page 2Linked to original sources

arXiv2026

Your AI-Generated Image Detector Can Secretly Achieve SOTA Accuracy, If Calibrated

Despite being trained on balanced datasets, existing AI-generated image detectors often exhibit systematic bias at test time, frequently misclassifying fake images as real. We hypothesize that this behavior stems from distributional shift in fake samples and implicit priors learned during training. Specifically, models tend to overfit to superficial artifacts that do not generalize well across different generation methods, leading to a misaligned decision threshold when faced with test-time distribution shift. To address this, we propose a theoretically grounded post-hoc calibration framework based on Bayesian decision theory. In particular, we introduce a learnable scalar correction to the model's logits, optimized on a small validation set from the target distribution while keeping the backbone frozen. This parametric adjustment compensates for distributional shift in model output, realigning the decision boundary even without requiring ground-truth labels. Experiments on challenging benchmarks show that our approach significantly improves robustness without retraining, offering a lightweight and principled solution for reliable and adaptive AI-generated image detection in the open world. Code is available at https://github.com/muliyangm/AIGI-Det-Calib.

cs.CV↗

arXiv2026

Towards Interpretable Hallucination Analysis and Mitigation in LVLMs via Contrastive Neuron Steering

LVLMs achieve remarkable multimodal understanding and generation but remain susceptible to hallucinations. Existing mitigation methods predominantly focus on output-level adjustments, leaving the internal mechanisms that give rise to these hallucinations largely unexplored. To gain a deeper understanding, we adopt a representation-level perspective by introducing sparse autoencoders (SAEs) to decompose dense visual embeddings into sparse, interpretable neurons. Through neuron-level analysis, we identify distinct neuron types, including always-on neurons and image-specific neurons. Our findings reveal that hallucinations often result from disruptions or spurious activations of image-specific neurons, while always-on neurons remain largely stable. Moreover, selectively enhancing or suppressing image-specific neurons enables controllable intervention in LVLM outputs, improving visual grounding and reducing hallucinations. Building on these insights, we propose Contrastive Neuron Steering (CNS), which identifies image-specific neurons via contrastive analysis between clean and noisy inputs. CNS selectively amplifies informative neurons while suppressing perturbation-induced activations, producing more robust and semantically grounded visual representations. This not only enhances visual understanding but also effectively mitigates hallucinations. By operating at the prefilling stage, CNS is fully compatible with existing decoding-stage methods. Extensive experiments on both hallucination-focused and general multimodal benchmarks demonstrate that CNS consistently reduces hallucinations while preserving overall multimodal understanding.

cs.CV↗

arXiv2025

A Turn Toward Better Alignment: Few-Shot Generative Adaptation with Equivariant Feature Rotation

Few-shot image generation aims to effectively adapt a source generative model to a target domain using very few training images. Most existing approaches introduce consistency constraints-typically through instance-level or distribution-level loss functions-to directly align the distribution patterns of source and target domains within their respective latent spaces. However, these strategies often fall short: overly strict constraints can amplify the negative effects of the domain gap, leading to distorted or uninformative content, while overly relaxed constraints may fail to leverage the source domain effectively. This limitation primarily stems from the inherent discrepancy in the underlying distribution structures of the source and target domains. The scarcity of target samples further compounds this issue by hindering accurate estimation of the target domain's distribution. To overcome these limitations, we propose Equivariant Feature Rotation (EFR), a novel adaptation strategy that aligns source and target domains at two complementary levels within a self-rotated proxy feature space. Specifically, we perform adaptive rotations within a parameterized Lie Group to transform both source and target features into an equivariant proxy space, where alignment is conducted. These learnable rotation matrices serve to bridge the domain gap by preserving intra-domain structural information without distortion, while the alignment optimization facilitates effective knowledge transfer from the source to the target domain. Comprehensive experiments on a variety of commonly used datasets demonstrate that our method significantly enhances the generative performance within the targeted domain.

cs.CV↗

arXiv2025

Towards Arbitrary Motion Completing via Hierarchical Continuous Representation

Physical motions are inherently continuous, and higher camera frame rates typically contribute to improved smoothness and temporal coherence. For the first time, we explore continuous representations of human motion sequences, featuring the ability to interpolate, inbetween, and even extrapolate any input motion sequences at arbitrary frame rates. To achieve this, we propose a novel parametric activation-induced hierarchical implicit representation framework, referred to as NAME, based on Implicit Neural Representations (INRs). Our method introduces a hierarchical temporal encoding mechanism that extracts features from motion sequences at multiple temporal scales, enabling effective capture of intricate temporal patterns. Additionally, we integrate a custom parametric activation function, powered by Fourier transformations, into the MLP-based decoder to enhance the expressiveness of the continuous representation. This parametric formulation significantly augments the model's ability to represent complex motion behaviors with high accuracy. Extensive evaluations across several benchmark datasets demonstrate the effectiveness and robustness of our proposed approach.

cs.CV↗

arXiv2025

Tempo as the Stable Cue: Hierarchical Mixture of Tempo and Beat Experts for Music to 3D Dance Generation

Music to 3D dance generation aims to synthesize realistic and rhythmically synchronized human dance from music. While existing methods often rely on additional genre labels to further improve dance generation, such labels are typically noisy, coarse, unavailable, or insufficient to capture the diversity of real-world music, which can result in rhythm misalignment or stylistic drift. In contrast, we observe that tempo, a core property reflecting musical rhythm and pace, remains relatively consistent across datasets and genres, typically ranging from 60 to 200 BPM. Based on this finding, we propose TempoMoE, a hierarchical tempo-aware Mixture-of-Experts module that enhances the diffusion model and its rhythm perception. TempoMoE organizes motion experts into tempo-structured groups for different tempo ranges, with multi-scale beat experts capturing fine- and long-range rhythmic dynamics. A Hierarchical Rhythm-Adaptive Routing dynamically selects and fuses experts from music features, enabling flexible, rhythm-aligned generation without manual genre labels. Extensive experiments demonstrate that TempoMoE achieves state-of-the-art results in dance quality and rhythm alignment.

cs.CV↗

arXiv2025

Revealing Perception and Generation Dynamics in LVLMs: Mitigating Hallucinations via Validated Dominance Correction

Large Vision-Language Models (LVLMs) have shown remarkable capabilities, yet hallucinations remain a persistent challenge. This work presents a systematic analysis of the internal evolution of visual perception and token generation in LVLMs, revealing two key patterns. First, perception follows a three-stage GATE process: early layers perform a Global scan, intermediate layers Approach and Tighten on core content, and later layers Explore supplementary regions. Second, generation exhibits an SAD (Subdominant Accumulation to Dominant) pattern, where hallucinated tokens arise from the repeated accumulation of subdominant tokens lacking support from attention (visual perception) or feed-forward network (internal knowledge). Guided by these findings, we devise the VDC (Validated Dominance Correction) strategy, which detects unsupported tokens and replaces them with validated dominant ones to improve output reliability. Extensive experiments across multiple models and benchmarks confirm that VDC substantially mitigates hallucinations.

cs.CV↗

arXiv2025

AStF: Motion Style Transfer via Adaptive Statistics Fusor

Human motion style transfer allows characters to appear less rigidity and more realism with specific style. Traditional arbitrary image style transfer typically process mean and variance which is proved effective. Meanwhile, similar methods have been adapted for motion style transfer. However, due to the fundamental differences between images and motion, relying on mean and variance is insufficient to fully capture the complex dynamic patterns and spatiotemporal coherence properties of motion data. Building upon this, our key insight is to bring two more coefficient, skewness and kurtosis, into the analysis of motion style. Specifically, we propose a novel Adaptive Statistics Fusor (AStF) which consists of Style Disentanglement Module (SDM) and High-Order Multi-Statistics Attention (HOS-Attn). We trained our AStF in conjunction with a Motion Consistency Regularization (MCR) discriminator. Experimental results show that, by providing a more comprehensive model of the spatiotemporal statistical patterns inherent in dynamic styles, our proposed AStF shows proficiency superiority in motion style transfers over state-of-the-arts. Our code and model are available at https://github.com/CHMimilanlan/AStF.

cs.CV↗

arXiv2025

From Physics to Machine Learning and Back: Part II - Learning and Observational Bias in PHM

Prognostics and Health Management ensures the reliability, safety, and efficiency of complex engineered systems by enabling fault detection, anticipating equipment failures, and optimizing maintenance activities throughout an asset lifecycle. However, real-world PHM presents persistent challenges: sensor data is often noisy or incomplete, available labels are limited, and degradation behaviors and system interdependencies can be highly complex and nonlinear. Physics-informed machine learning has emerged as a promising approach to address these limitations by embedding physical knowledge into data-driven models. This review examines how incorporating learning and observational biases through physics-informed modeling and data strategies can guide models toward physically consistent and reliable predictions. Learning biases embed physical constraints into model training through physics-informed loss functions and governing equations, or by incorporating properties like monotonicity. Observational biases influence data selection and synthesis to ensure models capture realistic system behavior through virtual sensing for estimating unmeasured states, physics-based simulation for data augmentation, and multi-sensor fusion strategies. The review then examines how these approaches enable the transition from passive prediction to active decision-making through reinforcement learning, which allows agents to learn maintenance policies that respect physical constraints while optimizing operational objectives. This closes the loop between model-based predictions, simulation, and actual system operation, empowering adaptive decision-making. Finally, the review addresses the critical challenge of scaling PHM solutions from individual assets to fleet-wide deployment. Fast adaptation methods including meta-learning and few-shot learning are reviewed alongside domain generalization techniques ...

cs.LG↗

arXiv2025

Deep Reinforcement Learning for Power Grid Multi-Stage Cascading Failure Mitigation

Cascading failures in power grids can lead to grid collapse, causing severe disruptions to social operations and economic activities. In certain cases, multi-stage cascading failures can occur. However, existing cascading-failure-mitigation strategies are usually single-stage-based, overlooking the complexity of the multi-stage scenario. This paper treats the multi-stage cascading failure problem as a reinforcement learning task and develops a simulation environment. The reinforcement learning agent is then trained via the deterministic policy gradient algorithm to achieve continuous actions. Finally, the effectiveness of the proposed approach is validated on the IEEE 14-bus and IEEE 118-bus systems.

cs.AI↗

arXiv2025

ChatGarment: Garment Estimation, Generation and Editing via Large Language Models

We introduce ChatGarment, a novel approach that leverages large vision-language models (VLMs) to automate the estimation, generation, and editing of 3D garments from images or text descriptions. Unlike previous methods that struggle in real-world scenarios or lack interactive editing capabilities, ChatGarment can estimate sewing patterns from in-the-wild images or sketches, generate them from text descriptions, and edit garments based on user instructions, all within an interactive dialogue. These sewing patterns can then be draped on a 3D body and animated. This is achieved by finetuning a VLM to directly generate a JSON file that includes both textual descriptions of garment types and styles, as well as continuous numerical attributes. This JSON file is then used to create sewing patterns through a programming parametric model. To support this, we refine the existing programming model, GarmentCode, by expanding its garment type coverage and simplifying its structure for efficient VLM fine-tuning. Additionally, we construct a large-scale dataset of image-to-sewing-pattern and text-to-sewing-pattern pairs through an automated data pipeline. Extensive evaluations demonstrate ChatGarment's ability to accurately reconstruct, generate, and edit garments from multimodal inputs, highlighting its potential to simplify workflows in fashion and gaming applications. Code and data are available at https://chatgarment.github.io/ .

cs.CV↗

arXiv2024

Exploiting Semantic Scene Reconstruction for Estimating Building Envelope Characteristics

Achieving the EU's climate neutrality goal requires retrofitting existing buildings to reduce energy use and emissions. A critical step in this process is the precise assessment of geometric building envelope characteristics to inform retrofitting decisions. Previous methods for estimating building characteristics, such as window-to-wall ratio, building footprint area, and the location of architectural elements, have primarily relied on applying deep-learning-based detection or segmentation techniques on 2D images. However, these approaches tend to focus on planar facade properties, limiting their accuracy and comprehensiveness when analyzing complete building envelopes in 3D. While neural scene representations have shown exceptional performance in indoor scene reconstruction, they remain under-explored for external building envelope analysis. This work addresses this gap by leveraging cutting-edge neural surface reconstruction techniques based on signed distance function (SDF) representations for 3D building analysis. We propose BuildNet3D, a novel framework to estimate geometric building characteristics from 2D image inputs. By integrating SDF-based representation with semantic modality, BuildNet3D recovers fine-grained 3D geometry and semantics of building envelopes, which are then used to automatically extract building characteristics. Our framework is evaluated on a range of complex building structures, demonstrating high accuracy and generalizability in estimating window-to-wall ratio and building footprint. The results underscore the effectiveness of BuildNet3D for practical applications in building analysis and retrofitting.

cs.CV↗

arXiv2024

GRADE: Generating Realistic And Dynamic Environments for Robotics Research with Isaac Sim

Synthetic data and novel rendering techniques have greatly influenced computer vision research in tasks like target tracking and human pose estimation. However, robotics research has lagged behind in leveraging it due to the limitations of most simulation frameworks, including the lack of low-level software control and flexibility, Robot Operating System integration, realistic physics, or photorealism. This hindered progress in (visual-)perception research, e.g. in autonomous robotics, especially in dynamic environments. Visual Simultaneous Localization and Mapping (V-SLAM), for instance, has been mostly developed passively, in static environments, and evaluated on few pre-recorded dynamic datasets due to the difficulties of realistically simulating dynamic worlds and the huge sim-to-real gap. To address these challenges, we present GRADE (Generating Realistic and Dynamic Environments), a highly customizable framework built upon NVIDIA Isaac Sim. We leverage Isaac's rendering capabilities and low-level APIs to populate and control the simulation, collect ground-truth data, and test online and offline approaches. Importantly, we introduce a new way to precisely repeat a recorded experiment within a physically enabled simulation while allowing environmental and simulation changes. Next, we collect a synthetic dataset of richly annotated videos in dynamic environments with a flying drone. Using that, we train detection and segmentation models for humans, closing the syn-to-real gap. Finally, we benchmark state-of-the-art dynamic V-SLAM algorithms, revealing their short tracking times and low generalization capabilities. We also show for the first time that the top-performing deep learning models do not achieve the best SLAM performance. Code and data are provided as open-source at https://grade.is.tue.mpg.de.

cs.RO↗

arXiv2024

DynaPix SLAM: A Pixel-Based Dynamic Visual SLAM Approach

Visual Simultaneous Localization and Mapping (V-SLAM) methods achieve remarkable performance in static environments, but face challenges in dynamic scenes where moving objects severely affect their core modules. To avoid this, dynamic V-SLAM approaches often leverage semantic information, geometric constraints, or optical flow. However, these methods are limited by imprecise estimations and their reliance on the accuracy of deep-learning models. Moreover, predefined thresholds for static/dynamic classification, the a-priori selection of dynamic object classes, and the inability to recognize unknown or unexpected moving objects, often degrade their performance. To address these limitations, we introduce DynaPix, a novel semantic-free V-SLAM system based on per-pixel motion probability estimation and an improved pose optimization process. The per-pixel motion probability is estimated using a static background differencing method on image data and optical flows computed on splatted frames. With DynaPix, we fully integrate these probabilities into map point selection and apply them through weighted bundle adjustment within the tracking and optimization modules of ORB-SLAM2. We thoroughly evaluate our method using the GRADE and TUM RGB-D datasets, showing significantly lower trajectory errors and longer tracking times in both static and dynamic sequences. The source code, datasets, and results are available at https://dynapix.is.tue.mpg.de/.

cs.RO↗

arXiv2023

The Non-zonal Rossby-Haurwitz Solutions of the 2D Euler Equations on a Rotating Ellipsoid

In this article, we investigate the incompressible 2D Euler equations on a rotating biaxial ellipsoid, which model the dynamics of the atmosphere of a Jovian planet. We study the non-zonal Rossby-Haurwitz solutions of the Euler equations on an ellipsoid, while previous works only considered the case of a sphere. Our main results include: the existence and uniqueness of the stationary Rossby-Haurwitz solutions; the construction of the traveling-wave solutions; and the demonstration of the Lyapunov instability of both the stationary and the traveling-wave solutions.

math.AP↗

arXiv2023

Learning from synthetic data generated with GRADE

Recently, synthetic data generation and realistic rendering has advanced tasks like target tracking and human pose estimation. Simulations for most robotics applications are obtained in (semi)static environments, with specific sensors and low visual fidelity. To solve this, we present a fully customizable framework for generating realistic animated dynamic environments (GRADE) for robotics research, first introduced in [1]. GRADE supports full simulation control, ROS integration, realistic physics, while being in an engine that produces high visual fidelity images and ground truth data. We use GRADE to generate a dataset focused on indoor dynamic scenes with people and flying objects. Using this, we evaluate the performance of YOLO and Mask R-CNN on the tasks of segmenting and detecting people. Our results provide evidence that using data generated with GRADE can improve the model performance when used for a pre-training step. We also show that, even training using only synthetic data, can generalize well to real-world images in the same application domain such as the ones from the TUM-RGBD dataset. The code, results, trained models, and the generated data are provided as open-source at https://eliabntt.github.io/grade-rr.

cs.CV↗

arXiv2023

Simulation of Dynamic Environments for SLAM

Simulation engines are widely adopted in robotics. However, they lack either full simulation control, ROS integration, realistic physics, or photorealism. Recently, synthetic data generation and realistic rendering has advanced tasks like target tracking and human pose estimation. However, when focusing on vision applications, there is usually a lack of information like sensor measurements or time continuity. On the other hand, simulations for most robotics tasks are performed in (semi)static environments, with specific sensors and low visual fidelity. To solve this, we introduced in our previous work a fully customizable framework for generating realistic animated dynamic environments (GRADE) [1]. We use GRADE to generate an indoor dynamic environment dataset and then compare multiple SLAM algorithms on different sequences. By doing that, we show how current research over-relies on known benchmarks, failing to generalize. Our tests with refined YOLO and Mask R-CNN models provide further evidence that additional research in dynamic SLAM is necessary. The code, results, and generated data are provided as open-source at https://eliabntt.github.io/grade-rrSimulation of Dynamic Environments for SLAM

cs.RO↗