arXiv subjects

Yue Ma

Publications and source records attributed to Yue Ma.

At least 127 records · Page 7Linked to original sources

arXiv2024

LVP-CLIP:Revisiting CLIP for Continual Learning with Label Vector Pool

Continual learning aims to update a model so that it can sequentially learn new tasks without forgetting previously acquired knowledge. Recent continual learning approaches often leverage the vision-language model CLIP for its high-dimensional feature space and cross-modality feature matching. Traditional CLIP-based classification methods identify the most similar text label for a test image by comparing their embeddings. However, these methods are sensitive to the quality of text phrases and less effective for classes lacking meaningful text labels. In this work, we rethink CLIP-based continual learning and introduce the concept of Label Vector Pool (LVP). LVP replaces text labels with training images as similarity references, eliminating the need for ideal text descriptions. We present three variations of LVP and evaluate their performance on class and domain incremental learning tasks. Leveraging CLIP's high dimensional feature space, LVP learning algorithms are task-order invariant. The new knowledge does not modify the old knowledge, hence, there is minimum forgetting. Different tasks can be learned independently and in parallel with low computational and memory demands. Experimental results show that proposed LVP-based methods outperform the current state-of-the-art baseline by a significant margin of 40.7%.

cs.CV↗

arXiv2024

InstantSwap: Fast Customized Concept Swapping across Sharp Shape Differences

Recent advances in Customized Concept Swapping (CCS) enable a text-to-image model to swap a concept in the source image with a customized target concept. However, the existing methods still face the challenges of inconsistency and inefficiency. They struggle to maintain consistency in both the foreground and background during concept swapping, especially when the shape difference is large between objects. Additionally, they either require time-consuming training processes or involve redundant calculations during inference. To tackle these issues, we introduce InstantSwap, a new CCS method that aims to handle sharp shape disparity at speed. Specifically, we first extract the bbox of the object in the source image automatically based on attention map analysis and leverage the bbox to achieve both foreground and background consistency. For background consistency, we remove the gradient outside the bbox during the swapping process so that the background is free from being modified. For foreground consistency, we employ a cross-attention mechanism to inject semantic information into both source and target concepts inside the box. This helps learn semantic-enhanced representations that encourage the swapping process to focus on the foreground objects. To improve swapping speed, we avoid computing gradients at each timestep but instead calculate them periodically to reduce the number of forward passes, which improves efficiency a lot with a little sacrifice on performance. Finally, we establish a benchmark dataset to facilitate comprehensive evaluation. Extensive evaluations demonstrate the superiority and versatility of InstantSwap. Project Page: https://instantswap.github.io/

cs.CV↗

arXiv2024

MagicStick: Controllable Video Editing via Control Handle Transformations

Text-based video editing has recently attracted considerable interest in changing the style or replacing the objects with a similar structure. Beyond this, we demonstrate that properties such as shape, size, location, motion, etc., can also be edited in videos. Our key insight is that the keyframe transformations of the specific internal feature (e.g., edge maps of objects or human pose), can easily propagate to other frames to provide generation guidance. We thus propose MagicStick, a controllable video editing method that edits the video properties by utilizing the transformation on the extracted internal control signals. In detail, to keep the appearance, we inflate both the pretrained image diffusion model and ControlNet to the temporal dimension and train low-rank adaptions (LORA) layers to fit the specific scenes. Then, in editing, we perform an inversion and editing framework. Differently, finetuned ControlNet is introduced in both inversion and generation for attention guidance with the proposed attention remix between the spatial attention maps of inversion and editing. Yet succinct, our method is the first method to show the ability of video property editing from the pre-trained text-to-image model. We present experiments on numerous examples within our unified framework. We also compare with shape-aware text-based editing and handcrafted motion video generation, demonstrating our superior temporal consistency and editing capability than previous works. The code and models are available on https://github.com/mayuelala/MagicStick.

cs.CV↗

arXiv2024

DiT4Edit: Diffusion Transformer for Image Editing

Despite recent advances in UNet-based image editing, methods for shape-aware object editing in high-resolution images are still lacking. Compared to UNet, Diffusion Transformers (DiT) demonstrate superior capabilities to effectively capture the long-range dependencies among patches, leading to higher-quality image generation. In this paper, we propose DiT4Edit, the first Diffusion Transformer-based image editing framework. Specifically, DiT4Edit uses the DPM-Solver inversion algorithm to obtain the inverted latents, reducing the number of steps compared to the DDIM inversion algorithm commonly used in UNet-based frameworks. Additionally, we design unified attention control and patches merging, tailored for transformer computation streams. This integration allows our framework to generate higher-quality edited images faster. Our design leverages the advantages of DiT, enabling it to surpass UNet structures in image editing, especially in high-resolution and arbitrary-size images. Extensive experiments demonstrate the strong performance of DiT4Edit across various editing scenarios, highlighting the potential of Diffusion Transformers in supporting image editing.

cs.CV↗

arXiv2024

Memoryless Multimodal Anomaly Detection via Student-Teacher Network and Signed Distance Learning

Unsupervised anomaly detection is a challenging computer vision task, in which 2D-based anomaly detection methods have been extensively studied. However, multimodal anomaly detection based on RGB images and 3D point clouds requires further investigation. The existing methods are mainly inspired by memory bank based methods commonly used in 2D-based anomaly detection, which may cost extra memory for storing mutimodal features. In present study, a novel memoryless method MDSS is proposed for multimodal anomaly detection, which employs a light-weighted student-teacher network and a signed distance function to learn from RGB images and 3D point clouds respectively, and complements the anomaly information from the two modalities. Specifically, a student-teacher network is trained with normal RGB images and masks generated from point clouds by a dynamic loss, and the anomaly score map could be obtained from the discrepancy between the output of student and teacher. Furthermore, the signed distance function learns from normal point clouds to predict the signed distances between points and surface, and the obtained signed distances are used to generate anomaly score map. Subsequently, the anomaly score maps are aligned to generate the final anomaly score map for detection. The experimental results indicate that MDSS is comparable but more stable than the SOTA memory bank based method Shape-guided, and furthermore performs better than other baseline methods.

cs.CV↗

arXiv2024

Follow-Your-Canvas: Higher-Resolution Video Outpainting with Extensive Content Generation

This paper explores higher-resolution video outpainting with extensive content generation. We point out common issues faced by existing methods when attempting to largely outpaint videos: the generation of low-quality content and limitations imposed by GPU memory. To address these challenges, we propose a diffusion-based method called \textit{Follow-Your-Canvas}. It builds upon two core designs. First, instead of employing the common practice of "single-shot" outpainting, we distribute the task across spatial windows and seamlessly merge them. It allows us to outpaint videos of any size and resolution without being constrained by GPU memory. Second, the source video and its relative positional relation are injected into the generation process of each window. It makes the generated spatial layout within each window harmonize with the source video. Coupling with these two designs enables us to generate higher-resolution outpainting videos with rich content while keeping spatial and temporal consistency. Follow-Your-Canvas excels in large-scale video outpainting, e.g., from 512X512 to 1152X2048 (9X), while producing high-quality and aesthetically pleasing results. It achieves the best quantitative results across various resolution and scale setups. The code is released on https://github.com/mayuelala/FollowYourCanvas

cs.CV↗

arXiv2024

Decorrelated forward regression for high dimensional data analysis

Forward regression is a crucial methodology for automatically identifying important predictors from a large pool of potential covariates. In contexts with moderate predictor correlation, forward selection techniques can achieve screening consistency. However, this property gradually becomes invalid in the presence of substantially correlated variables, especially in high-dimensional datasets where strong correlations exist among predictors. This dilemma is encountered by other model selection methods in literature as well. To address these challenges, we introduce a novel decorrelated forward (DF) selection framework for generalized mean regression models, including prevalent models, such as linear, logistic, Poisson, and quasi likelihood. The DF selection framework stands out because of its ability to convert generalized mean regression models into linear ones, thus providing a clear interpretation of the forward selection process. It also offers a closed-form expression for forward iteration, to improve practical applicability and efficiency. Theoretically, we establish the screening consistency of DF selection and determine the upper bound of the selected submodel's size. To reduce computational burden, we develop a thresholding DF algorithm that provides a stopping rule for the forward-searching process. Simulations and two real data applications show the outstanding performance of our method compared with some existing model selection methods.

stat.ME↗

arXiv2024

Freehand Sketch Generation from Mechanical Components

Drawing freehand sketches of mechanical components on multimedia devices for AI-based engineering modeling has become a new trend. However, its development is being impeded because existing works cannot produce suitable sketches for data-driven research. These works either generate sketches lacking a freehand style or utilize generative models not originally designed for this task resulting in poor effectiveness. To address this issue, we design a two-stage generative framework mimicking the human sketching behavior pattern, called MSFormer, which is the first time to produce humanoid freehand sketches tailored for mechanical components. The first stage employs Open CASCADE technology to obtain multi-view contour sketches from mechanical components, filtering perturbing signals for the ensuing generation process. Meanwhile, we design a view selector to simulate viewpoint selection tasks during human sketching for picking out information-rich sketches. The second stage translates contour sketches into freehand sketches by a transformer-based generator. To retain essential modeling features as much as possible and rationalize stroke distribution, we introduce a novel edge-constraint stroke initialization. Furthermore, we utilize a CLIP vision encoder and a new loss function incorporating the Hausdorff distance to enhance the generalizability and robustness of the model. Extensive experiments demonstrate that our approach achieves state-of-the-art performance for generating freehand sketches in the mechanical domain. Project page: https://mcfreeskegen.github.io .

cs.CV↗

arXiv2024

Developing Smart MAVs for Autonomous Inspection in GPS-denied Constructions

Smart Micro Aerial Vehicles (MAVs) have transformed infrastructure inspection by enabling efficient, high-resolution monitoring at various stages of construction, including hard-to-reach areas. Traditional manual operation of drones in GPS-denied environments, such as industrial facilities and infrastructure, is labour-intensive, tedious and prone to error. This study presents an innovative framework for smart MAV inspections in such complex and GPS-denied indoor environments. The framework features a hierarchical perception and planning system that identifies regions of interest and optimises task paths. It also presents an advanced MAV system with enhanced localisation and motion planning capabilities, integrated with Neural Reconstruction technology for comprehensive 3D reconstruction of building structures. The effectiveness of the framework was empirically validated in a 4,000 square meters indoor infrastructure facility with an interior length of 80 metres, a width of 50 metres and a height of 7 metres. The main structure consists of columns and walls. Experimental results show that our MAV system performs exceptionally well in autonomous inspection tasks, achieving a 100\% success rate in generating and executing scan paths. Extensive experiments validate the manoeuvrability of our developed MAV, achieving a 100\% success rate in motion planning with a tracking error of less than 0.1 metres. In addition, the enhanced reconstruction method using 3D Gaussian Splatting technology enables the generation of high-fidelity rendering models from the acquired data. Overall, our novel method represents a significant advancement in the use of robotics for infrastructure inspection.

cs.RO↗

arXiv2024

Feasibility of Liquid-phase Xenon Proportional Scintillation for Low-energy Physics

Dual phase xenon time projection chambers (TPCs) detect both the scintillation photons and ionization electrons created by energy depositions within the liquid xenon (LXe) volume. The electrons are extracted from the interaction site through a gas gap, where they meet a high electric field where proportional scintillation occurs. This converts the electron signal into a light signal, and yields a high electron detection efficiency with a gain of tens of photoelectrons (PE) per electron. This technique of detecting both scintillation and ionization gives dual phase xenon TPCs the capability to distinguish between electronic and nuclear recoils, which is a key part of how these detectors are able to reach world-leading limits on Weakly Interacting Massive Particle (WIMP) dark matter. However, not all electrons can be extracted through the liquid-gas interface, and a constant millimeter-scale gas gap needs to be maintained, which may be a technological challenge if dual-phase xenon TPCs are to be scaled up for future dark matter searches. Furthermore, there is a background of single-electron peaks that follow a large ionization signal (S2) of unclear origin which may be due in part to the liquid-gas interface, and limits the sensitivity of these detectors towards low mass dark matter. In this paper, we demonstrate that a purely single-phase liquid xenon TPC which produces proportional scintillation directly in the liquid is still capable of discriminating between electronic and nuclear recoils, but that the background of single-electrons following an S2 is still likely unrelated to the liquid-gas interface.

physics.ins-det↗

arXiv2024

Alleviating Over-Smoothing via Aggregation over Compact Manifolds

Graph neural networks (GNNs) have achieved significant success in various applications. Most GNNs learn the node features with information aggregation of its neighbors and feature transformation in each layer. However, the node features become indistinguishable after many layers, leading to performance deterioration: a significant limitation known as over-smoothing. Past work adopted various techniques for addressing this issue, such as normalization and skip-connection of layer-wise output. After the study, we found that the information aggregations in existing work are all contracted aggregations, with the intrinsic property that features will inevitably converge to the same single point after many layers. To this end, we propose the aggregation over compacted manifolds method (ACM) that replaces the existing information aggregation with aggregation over compact manifolds, a special type of manifold, which avoids contracted aggregations. In this work, we theoretically analyze contracted aggregation and its properties. We also provide an extensive empirical evaluation that shows ACM can effectively alleviate over-smoothing and outperforms the state-of-the-art. The code can be found in https://github.com/DongzhuoranZhou/ACM.git.

cs.LG↗

arXiv2024

Follow-Your-Emoji: Fine-Controllable and Expressive Freestyle Portrait Animation

We present Follow-Your-Emoji, a diffusion-based framework for portrait animation, which animates a reference portrait with target landmark sequences. The main challenge of portrait animation is to preserve the identity of the reference portrait and transfer the target expression to this portrait while maintaining temporal consistency and fidelity. To address these challenges, Follow-Your-Emoji equipped the powerful Stable Diffusion model with two well-designed technologies. Specifically, we first adopt a new explicit motion signal, namely expression-aware landmark, to guide the animation process. We discover this landmark can not only ensure the accurate motion alignment between the reference portrait and target motion during inference but also increase the ability to portray exaggerated expressions (i.e., large pupil movements) and avoid identity leakage. Then, we propose a facial fine-grained loss to improve the model's ability of subtle expression perception and reference portrait appearance reconstruction by using both expression and facial masks. Accordingly, our method demonstrates significant performance in controlling the expression of freestyle portraits, including real humans, cartoons, sculptures, and even animals. By leveraging a simple and effective progressive generation strategy, we extend our model to stable long-term animation, thus increasing its potential application value. To address the lack of a benchmark for this field, we introduce EmojiBench, a comprehensive benchmark comprising diverse portrait images, driving videos, and landmarks. We show extensive evaluations on EmojiBench to verify the superiority of Follow-Your-Emoji.

cs.CV↗

arXiv2024

A rigidity property for a type of wave-Klein-Gordon system

In this paper we investigate the rigidity property of a wave component coupled in a wave-Klein-Gordon system. We prove that when the radiation field of the wave component vanishes at the null infinity, the initial data of this component also vanish, therefor there is no wave in the whole spacetime

math.AP↗

arXiv2024

The Euclidean-hyperboloidal foliation method. Application to f(R) modified gravity

This paper is a part of a series devoted to the Euclidean-hyperboloidal foliation method introduced by the authors for investigating the global existence problem associated with nonlinear systems of coupled wave-Klein-Gordon equations with small data. This method was developed especially for investigating the initial value problem for the Einstein-massive field system in wave gauge. Here, we study the (fourth-order) field equations of f(R) modified gravity and investigate the global dynamical behavior of the gravitational field in the near-Minkowski regime. We establish the existence of a globally hyperbolic Cauchy development approaching Minkowski spacetime (in spacelike, null, and timelike directions), when the initial data set is sufficiently close to an asymptotically Euclidean and spacelike hypersurface in Minkowski spacetime. We cast the (fourth-order) f(R)-field equations in the form of a second-order wave-Klein-Gordon system, which has an analogous structure to the Einstein-massive field system but, in addition, involves a (possibly small) effective mass parameter. We establish the nonlinear stability of the Minkowski spacetime in the context of f(R) gravity, when the integrand f(R) in the action functional can be taken to be arbitrarily close to the integrand R of the standard Hilbert-Einstein action.

gr-qc↗

arXiv2024

Heterogeneous Network Based Contrastive Learning Method for PolSAR Land Cover Classification

Polarimetric synthetic aperture radar (PolSAR) image interpretation is widely used in various fields. Recently, deep learning has made significant progress in PolSAR image classification. Supervised learning (SL) requires a large amount of labeled PolSAR data with high quality to achieve better performance, however, manually labeled data is insufficient. This causes the SL to fail into overfitting and degrades its generalization performance. Furthermore, the scattering confusion problem is also a significant challenge that attracts more attention. To solve these problems, this article proposes a Heterogeneous Network based Contrastive Learning method(HCLNet). It aims to learn high-level representation from unlabeled PolSAR data for few-shot classification according to multi-features and superpixels. Beyond the conventional CL, HCLNet introduces the heterogeneous architecture for the first time to utilize heterogeneous PolSAR features better. And it develops two easy-to-use plugins to narrow the domain gap between optics and PolSAR, including feature filter and superpixel-based instance discrimination, which the former is used to enhance the complementarity of multi-features, and the latter is used to increase the diversity of negative samples. Experiments demonstrate the superiority of HCLNet on three widely used PolSAR benchmark datasets compared with state-of-the-art methods. Ablation studies also verify the importance of each component. Besides, this work has implications for how to efficiently utilize the multi-features of PolSAR data to learn better high-level representation in CL and how to construct networks suitable for PolSAR data better.

cs.CV↗

arXiv2024

Follow-Your-Click: Open-domain Regional Image Animation via Short Prompts

Despite recent advances in image-to-video generation, better controllability and local animation are less explored. Most existing image-to-video methods are not locally aware and tend to move the entire scene. However, human artists may need to control the movement of different objects or regions. Additionally, current I2V methods require users not only to describe the target motion but also to provide redundant detailed descriptions of frame contents. These two issues hinder the practical utilization of current I2V tools. In this paper, we propose a practical framework, named Follow-Your-Click, to achieve image animation with a simple user click (for specifying what to move) and a short motion prompt (for specifying how to move). Technically, we propose the first-frame masking strategy, which significantly improves the video generation quality, and a motion-augmented module equipped with a short motion prompt dataset to improve the short prompt following abilities of our model. To further control the motion speed, we propose flow-based motion magnitude control to control the speed of target movement more precisely. Our framework has simpler yet precise user control and better generation performance than previous methods. Extensive experiments compared with 7 baselines, including both commercial tools and research methods on 8 metrics, suggest the superiority of our approach. Project Page: https://follow-your-click.github.io/

cs.CV↗

arXiv2024

Nonlinear stability of self-gravitating massive fields

We consider the global evolution problem for Einstein's field equations in the near-Minkowski regime and study the long-time dynamics of a massive scalar field evolving under its own gravitational field. We establish the existence of a globally hyperbolic Cauchy development associated with any initial data set that is sufficiently close to a data set in Minkowski spacetime. In addition to applying to massive fields, our theory allows us to cover metrics with slow decay in space. The strategy of proof, proposed here and referred to as the Euclidean-Hyperboloidal Foliation Method, applies, more generally, to nonlinear systems of coupled wave and Klein-Gordon equations. It is based on a spacetime foliation defined by merging together asymptotically Euclidean hypersurfaces (covering spacelike infinity) and asymptotically hyperboloidal hypersurfaces (covering timelike infinity). A transition domain (reaching null infinity) limited by two asymptotic light cones is introduced in order to realize this merging. On the one hand, we exhibit a boost-rotation hierarchy property (as we call it) which is associated with Minkowski's Killing fields and is enjoyed by commutators of curved wave operators and, on the other hand, we exhibit a metric hierarchy property (as we call it) enjoyed by components of Einstein's field equations in frames associated with our Euclidean-hyperboloidal foliation. The core of the argument is, on the one hand, the derivation of novel integral and pointwise estimates which lead us to almost sharp decay properties (at timelike, null, and spacelike infinity) and, on the other hand, the control of the (quasi-linear and semi-linear) coupling between the geometric and matter parts of the Einstein equations.

gr-qc↗

arXiv2024

Limitations of probabilistic error cancellation for open dynamics beyond sampling overhead

Quantum simulation of dynamics is an important goal in the NISQ era, within which quantum error mitigation may be a viable path towards modifying or eliminating the effects of noise. Most studies on quantum error mitigation have been focused on the resource cost due to its exponential scaling in the circuit depth. Methods such as probabilistic error cancellation rely on discretizing the evolution into finite time steps and applying the mitigation layer after each time step, modifying only the noise part without any Hamiltonian-dependence. This may lead to Trotter-like errors in the simulation results even if the error mitigation is implemented ideally, which means that the number of samples is taken as infinite. Here we analyze the aforementioned errors which have been largely neglected before. We show that, they are determined by the commutating relations between the superoperators of the unitary part, the device noise part and the noise part of the open dynamics to be simulated. We include both digital quantum simulation and analog quantum simulation setups, and consider defining the ideal error mitigation map both by exactly inverting the noise channel and by approximating it to the first order in the time step. We take single-qubit toy models to numerically demonstrate our findings. Our results illustrate fundamental limitations of applying probabilistic error cancellation in a stepwise manner to continuous dynamics, thus motivating the investigations of truly time-continuous error cancellation methods.

quant-ph↗