arXiv subjects

Wen Yang

Publications and source records attributed to Wen Yang.

At least 109 records · Page 6Linked to original sources

arXiv2024

Robust Tiny Object Detection in Aerial Images amidst Label Noise

Precise detection of tiny objects in remote sensing imagery remains a significant challenge due to their limited visual information and frequent occurrence within scenes. This challenge is further exacerbated by the practical burden and inherent errors associated with manual annotation: annotating tiny objects is laborious and prone to errors (i.e., label noise). Training detectors for such objects using noisy labels often leads to suboptimal performance, with networks tending to overfit on noisy labels. In this study, we address the intricate issue of tiny object detection under noisy label supervision. We systematically investigate the impact of various types of noise on network training, revealing the vulnerability of object detectors to class shifts and inaccurate bounding boxes for tiny objects. To mitigate these challenges, we propose a DeNoising Tiny Object Detector (DN-TOD), which incorporates a Class-aware Label Correction (CLC) scheme to address class shifts and a Trend-guided Learning Strategy (TLS) to handle bounding box noise. CLC mitigates inaccurate class supervision by identifying and filtering out class-shifted positive samples, while TLS reduces noisy box-induced erroneous supervision through sample reweighting and bounding box regeneration. Additionally, Our method can be seamlessly integrated into both one-stage and two-stage object detection pipelines. Comprehensive experiments conducted on synthetic (i.e., noisy AI-TOD-v2.0 and DOTA-v2.0) and real-world (i.e., AI-TOD) noisy datasets demonstrate the robustness of DN-TOD under various types of label noise. Notably, when applied to the strong baseline RFLA, DN-TOD exhibits a noteworthy performance improvement of 4.9 points under 40% mixed noise. Datasets, codes, and models will be made publicly available.

cs.CV↗

arXiv2024

Existence and nonexistence of minimizer for Thomas-Fermi-Dirac-Von Weizsäcker model on lattice graph

The focus of our paper is to investigate the possibility of a minimizer for the Thomas-Fermi-Dirac-von Weizsäcker model on the lattice graph $\mathbb{Z}^{3}$. The model is described by the following functional: \begin{equation*} E(φ)=\sum_{y\in\mathbb{Z}^{3}}\left(|\nablaφ(y)|^2+ (φ(y))^{\frac{10}{3}}-(φ(y))^{\frac{8}{3}}\right)+ \sum_{x,y\in\mathbb{Z}^{3}\atop ~\ y\neq x\hfill}\frac{φ^2(x)φ^2(y)}{|x-y|}, \end{equation*} with the additional constraint that $\sum\limits_{y\in\mathbb{Z}^{3}} φ^2(y)=m$ is sufficiently small. We also prove the nonexistence of a minimizer provided the mass $m$ is adequately large. Furthermore, we extend our analysis to a subset $Ω\subset \mathbb{Z}^{3}$ and prove the nonexistence of a minimizer for the following functional: \begin{equation*} E(Ω)=|\partialΩ|+\sum_{x,y\inΩ\atop ~y\neq x\hfill}\frac{1}{|x-y|}, \end{equation*} under the constraint that $|Ω|=V$ is sufficiently large.

math.AP↗

arXiv2023

Lump type solutions: Backlund transformation and spectral properties

There are various different ways to obtain traveling waves of lump type for the KP equation. We propose a general and simple approach to derive them via a Backlund transformation. This enables us to establish an explicit connection between those low energy solutions and high energy ones. Based on this construction, spectral analysis of the degree $6$ solutions is then carried out in details. The analysis of higher energy ones can be done in an inductive way.

math.AP↗

arXiv2023

BigTranslate: Augmenting Large Language Models with Multilingual Translation Capability over 100 Languages

Large language models (LLMs) demonstrate promising translation performance among various natural languages. However, many LLMs especially the open-sourced ones, such as BLOOM and LLaMA, are English-dominant and support only dozens of natural languages, making the potential of LLMs on language translation less explored. In this work, we present BigTranslate which adapts LLaMA that covers only 20 languages and enhances it with multilingual translation capability on more than 100 languages. BigTranslate is built upon LLaMA-13B and it is optimized in three steps. First, we continue training LLaMA with massive Chinese monolingual data. Second, we continue training the model with a large-scale parallel dataset that covers 102 natural languages. Third, we instruct-tune the foundation model with multilingual translation instructions, leading to our BigTranslate model. The preliminary experiments on multilingual translation show that BigTranslate performs comparably with ChatGPT and Google Translate in many languages and even outperforms ChatGPT in 8 language pairs. We release the BigTranslate model and hope it can advance the research progress.

cs.CL↗

arXiv2023

Rethinking Scale Imbalance in Semi-supervised Object Detection for Aerial Images

This paper focuses on the scale imbalance problem of semi-supervised object detection(SSOD) in aerial images. Compared to natural images, objects in aerial images show smaller sizes and larger quantities per image, increasing the difficulty of manual annotation. Meanwhile, the advanced SSOD technique can train superior detectors by leveraging limited labeled data and massive unlabeled data, saving annotation costs. However, as an understudied task in aerial images, SSOD suffers from a drastic performance drop when facing a large proportion of small objects. By analyzing the predictions between small and large objects, we identify three imbalance issues caused by the scale bias, i.e., pseudo-label imbalance, label assignment imbalance, and negative learning imbalance. To tackle these issues, we propose a novel Scale-discriminative Semi-Supervised Object Detection (S^3OD) learning pipeline for aerial images. In our S^3OD, three key components, Size-aware Adaptive Thresholding (SAT), Size-rebalanced Label Assignment (SLA), and Teacher-guided Negative Learning (TNL), are proposed to warrant scale unbiased learning. Specifically, SAT adaptively selects appropriate thresholds to filter pseudo-labels for objects at different scales. SLA balances positive samples of objects at different scales through resampling and reweighting. TNL alleviates the imbalance in negative samples by leveraging information generated by a teacher model. Extensive experiments conducted on the DOTA-v1.5 benchmark demonstrate the superiority of our proposed methods over state-of-the-art competitors. Codes will be released soon.

cs.CV↗

arXiv2023

Uniqueness of lump solutions of KP-I equation

The KP-I equation has family of solutions which decay to zero at space infinity. One of these solutions is the classical lump solution. This is a traveling wave, and the KP-I equation in this case reduces to the Boussinesq equation. In this paper we classify the lump type solutions of the Boussinesq equation. Using a robust inverse scattering transform developed by Bilman-Miller, we show that the lump type solutions are rational and their tau function has to be a polynomial of degree $k(k+1)$. In particular, this implies that the lump solution is the unique ground state of the KP-I equation (as conjectured by Klein and Saut in \cite{Klein0}). Our result generalizes a theorem by Airault-McKean-Moser on the classification of rational solutions for the KdV equation.

math.AP↗

arXiv2023

CrossZoom: Simultaneously Motion Deblurring and Event Super-Resolving

Even though the collaboration between traditional and neuromorphic event cameras brings prosperity to frame-event based vision applications, the performance is still confined by the resolution gap crossing two modalities in both spatial and temporal domains. This paper is devoted to bridging the gap by increasing the temporal resolution for images, i.e., motion deblurring, and the spatial resolution for events, i.e., event super-resolving, respectively. To this end, we introduce CrossZoom, a novel unified neural Network (CZ-Net) to jointly recover sharp latent sequences within the exposure period of a blurry input and the corresponding High-Resolution (HR) events. Specifically, we present a multi-scale blur-event fusion architecture that leverages the scale-variant properties and effectively fuses cross-modality information to achieve cross-enhancement. Attention-based adaptive enhancement and cross-interaction prediction modules are devised to alleviate the distortions inherent in Low-Resolution (LR) events and enhance the final results through the prior blur-event complementary information. Furthermore, we propose a new dataset containing HR sharp-blurry images and the corresponding HR-LR event streams to facilitate future research. Extensive qualitative and quantitative experiments on synthetic and real-world datasets demonstrate the effectiveness and robustness of the proposed method. Codes and datasets are released at https://bestrivenzc.github.io/CZ-Net/.

cs.CV↗

arXiv2023

Geometric density of states of electronic structures for local responses: Phase information from the amplitudes of STM measurement

Electronic band structures underlie the physical properties of crystalline materials, their geometrical exploration renovates the conventional cognition and brings about novel applications. Inspired by geometry phases, we introduce a geometric amplitude named as the geometric density of states (GDOS) dictated by the differential curvature of the constant-energy contour. The GDOS determines the amplitude of the real-space Green's function making it attain the ultimate expression with transparent physics. The local responses of crystalline materials are usually formulated by the real-space Green's function, so the relevant physics should be refreshed by GDOS. As an example of local responses, we suggest using scanning tunneling microscopy (STM) to characterize the surface states of three-dimensional topological insulator under an in-plane magnetic field. The GDOS favors the straightforward simulation of STM measurement without resorting to Fourier transform of the real-space measurement, and also excavates the unexplored potential of STM measurement to extract the phase information of wavefunction through its amplitude, i.e., the spin and curvature textures. Therefore, the proposed GDOS deepens the understanding of electronic band structures and is indispensable in local responses, and it should be universal for any periodic systems.

cond-mat.mes-hall↗

arXiv2023

QuadricsNet: Learning Concise Representation for Geometric Primitives in Point Clouds

This paper presents a novel framework to learn a concise geometric primitive representation for 3D point clouds. Different from representing each type of primitive individually, we focus on the challenging problem of how to achieve a concise and uniform representation robustly. We employ quadrics to represent diverse primitives with only 10 parameters and propose the first end-to-end learning-based framework, namely QuadricsNet, to parse quadrics in point clouds. The relationships between quadrics mathematical formulation and geometric attributes, including the type, scale and pose, are insightfully integrated for effective supervision of QuaidricsNet. Besides, a novel pattern-comprehensive dataset with quadrics segments and objects is collected for training and evaluation. Experiments demonstrate the effectiveness of our concise representation and the robustness of QuadricsNet. Our code is available at \url{https://github.com/MichaelWu99-lab/QuadricsNet}

cs.RO↗

arXiv2023

2D-3D Pose Tracking with Multi-View Constraints

Camera localization in 3D LiDAR maps has gained increasing attention due to its promising ability to handle complex scenarios, surpassing the limitations of visual-only localization methods. However, existing methods mostly focus on addressing the cross-modal gaps, estimating camera poses frame by frame without considering the relationship between adjacent frames, which makes the pose tracking unstable. To alleviate this, we propose to couple the 2D-3D correspondences between adjacent frames using the 2D-2D feature matching, establishing the multi-view geometrical constraints for simultaneously estimating multiple camera poses. Specifically, we propose a new 2D-3D pose tracking framework, which consists: a front-end hybrid flow estimation network for consecutive frames and a back-end pose optimization module. We further design a cross-modal consistency-based loss to incorporate the multi-view constraints during the training and inference process. We evaluate our proposed framework on the KITTI and Argoverse datasets. Experimental results demonstrate its superior performance compared to existing frame-by-frame 2D-3D pose tracking methods and state-of-the-art vision-only pose tracking algorithms. More online pose tracking videos are available at \url{https://youtu.be/yfBRdg7gw5M}

cs.RO↗

arXiv2023

Generalizing Event-Based Motion Deblurring in Real-World Scenarios

Event-based motion deblurring has shown promising results by exploiting low-latency events. However, current approaches are limited in their practical usage, as they assume the same spatial resolution of inputs and specific blurriness distributions. This work addresses these limitations and aims to generalize the performance of event-based deblurring in real-world scenarios. We propose a scale-aware network that allows flexible input spatial scales and enables learning from different temporal scales of motion blur. A two-stage self-supervised learning scheme is then developed to fit real-world data distribution. By utilizing the relativity of blurriness, our approach efficiently ensures the restored brightness and structure of latent images and further generalizes deblurring performance to handle varying spatial and temporal scales of motion blur in a self-distillation manner. Our method is extensively evaluated, demonstrating remarkable performance, and we also introduce a real-world dataset consisting of multi-scale blurry frames and events to facilitate research in event-based deblurring.

cs.CV↗

arXiv2023

Motion Robust High-Speed Light-Weighted Object Detection With Event Camera

In this work, we propose a motion robust and high-speed detection pipeline which better leverages the event data. First, we design an event stream representation called temporal active focus (TAF), which efficiently utilizes the spatial-temporal asynchronous event stream, constructing event tensors robust to object motions. Then, we propose a module called the bifurcated folding module (BFM), which encodes the rich temporal information in the TAF tensor at the input layer of the detector. Following this, we design a high-speed lightweight detector called agile event detector (AED) plus a simple but effective data augmentation method, to enhance the detection accuracy and reduce the model's parameter. Experiments on two typical real-scene event camera object detection datasets show that our method is competitive in terms of accuracy, efficiency, and the number of parameters. By classifying objects into multiple motion levels based on the optical flow density metric, we further illustrated the robustness of our method for objects with different velocities relative to the camera. The codes and trained models are available at https://github.com/HarmoniaLeo/FRLW-EvD .

cs.CV↗

arXiv2023

Liouville type theorems and periodic solutions for the nonhomogeneous parabolic systems

In the present paper we derive Liouville type results and existence of periodic solutions for $χ^{(2)}$ type systems with non-homogeneous nonlinearities. Moreover, we prove both universal bounds as well as singularity and decay estimates for this class of problems. In this study, we have to face new difficulties due to the non-homogenous nonlinearities. To overcome this issue, we carry out delicate integral estimates for this class of nonlinearities and modify the usual scaling and blow up arguments. This seems to be the first result for parabolic systems with non-homogeneous nonlinearities.

math.AP↗

arXiv2023

Tunneling valley Hall effect driven by tilted Dirac fermions

Valleytronics is a research field utilizing a valley degree of freedom of electrons for information processing and storage. A strong valley polarization is critical for realistic valleytronic applications. Here, we predict a tunneling valley Hall effect (TVHE) driven by tilted Dirac fermions in all-in-one tunnel junctions based on a two-dimensional (2D) valley material. Different doping of the electrode and spacer regions in these tunnel junctions results in momentum filtering of the tunneling Dirac fermions, generating a strong transverse valley Hall current dependent on the Dirac-cone tilting. Using the parameters of an existing 2D valley material, we demonstrate that such a TVHE is much stronger than that induced by the intrinsic Berry curvature mechanism reported previously. Finally, we predict that resonant tunneling can occur in a tunnel junction with properly engineered device parameters such as the spacer width and transport direction, providing significant enhancement of the valley Hall angle. Our work opens a new approach to generate valley polarization in realistic valleytronic systems.

cond-mat.mes-hall↗

arXiv2023

Affine Toda system of $\mathbf{A}$ and $\mathbf{C}^t$ type: compactness and affine Weyl group

The local mass is a fundamental quantized information that characterizes the blow-up solution to the Toda system and has a profound relationship with its underlying algebraic structure. In \cite{Lin-Yang-Zhong-2020}, it was observed that the associated Weyl group can be employed to represent this information for the $\mathbf{A}_n$, $\mathbf{B}_n$, $\mathbf{C}_n$ and $\mathbf{G}_2$ type Toda systems. The relationship between the local mass of blow-up solution and the corresponding affine Weyl group is further explored for some affine $\mathbf{B}$ type Toda systems in \cite{Cui-Wei-Yang-Zhang-2022}, where the possible local masses are explicitly expressed in terms of $8$ types. The current work presents a comprehensive study of the general affine $\mathbf{A}$ and $\mathbf{C}^t$ type Toda systems with arbitrary rank. At each stage of the blow-up process (via scaling), we can employ certain elements (known as "set chains") in the corresponding affine Weyl group to measure the variation of local mass. Consequently, we obtain the a priori estimate of the affine $\mathbf{A}$ and $\mathbf{C}^t$ type Toda systems with arbitrary number of singularities.

math.AP↗

arXiv2023

Self-Supervised Scene Dynamic Recovery from Rolling Shutter Images and Events

Scene Dynamic Recovery (SDR) by inverting distorted Rolling Shutter (RS) images to an undistorted high frame-rate Global Shutter (GS) video is a severely ill-posed problem due to the missing temporal dynamic information in both RS intra-frame scanlines and inter-frame exposures, particularly when prior knowledge about camera/object motions is unavailable. Commonly used artificial assumptions on scenes/motions and data-specific characteristics are prone to producing sub-optimal solutions in real-world scenarios. To address this challenge, we propose an event-based SDR network within a self-supervised learning paradigm, i.e., SelfUnroll. We leverage the extremely high temporal resolution of event cameras to provide accurate inter/intra-frame dynamic information. Specifically, an Event-based Inter/intra-frame Compensator (E-IC) is proposed to predict the per-pixel dynamic between arbitrary time intervals, including the temporal transition and spatial translation. Exploring connections in terms of RS-RS, RS-GS, and GS-RS, we explicitly formulate mutual constraints with the proposed E-IC, resulting in supervisions without ground-truth GS images. Extensive evaluations over synthetic and real datasets demonstrate that the proposed method achieves state-of-the-art and shows remarkable performance for event-based RS2GS inversion in real-world scenarios. The dataset and code are available at https://w3un.github.io/selfunroll/.

cs.CV↗

arXiv2023

Dynamic Coarse-to-Fine Learning for Oriented Tiny Object Detection

Detecting arbitrarily oriented tiny objects poses intense challenges to existing detectors, especially for label assignment. Despite the exploration of adaptive label assignment in recent oriented object detectors, the extreme geometry shape and limited feature of oriented tiny objects still induce severe mismatch and imbalance issues. Specifically, the position prior, positive sample feature, and instance are mismatched, and the learning of extreme-shaped objects is biased and unbalanced due to little proper feature supervision. To tackle these issues, we propose a dynamic prior along with the coarse-to-fine assigner, dubbed DCFL. For one thing, we model the prior, label assignment, and object representation all in a dynamic manner to alleviate the mismatch issue. For another, we leverage the coarse prior matching and finer posterior constraint to dynamically assign labels, providing appropriate and relatively balanced supervision for diverse instances. Extensive experiments on six datasets show substantial improvements to the baseline. Notably, we obtain the state-of-the-art performance for one-stage detectors on the DOTA-v1.5, DOTA-v2.0, and DIOR-R datasets under single-scale training and testing. Codes are available at https://github.com/Chasel-Tsui/mmrotate-dcfl.

cs.CV↗

arXiv2023

Recovering Continuous Scene Dynamics from A Single Blurry Image with Events

This paper aims at demystifying a single motion-blurred image with events and revealing temporally continuous scene dynamics encrypted behind motion blurs. To achieve this end, an Implicit Video Function (IVF) is learned to represent a single motion blurred image with concurrent events, enabling the latent sharp image restoration of arbitrary timestamps in the range of imaging exposures. Specifically, a dual attention transformer is proposed to efficiently leverage merits from both modalities, i.e., the high temporal resolution of event features and the smoothness of image features, alleviating temporal ambiguities while suppressing the event noise. The proposed network is trained only with the supervision of ground-truth images of limited referenced timestamps. Motion- and texture-guided supervisions are employed simultaneously to enhance restorations of the non-referenced timestamps and improve the overall sharpness. Experiments on synthetic, semi-synthetic, and real-world datasets demonstrate that our proposed method outperforms state-of-the-art methods by a large margin in terms of both objective PSNR and SSIM measurements and subjective evaluations.

cs.CV↗