arXiv subjects

Yanjie Li

Publications and source records attributed to Yanjie Li.

At least 55 records · Page 3Linked to original sources

arXiv2024

PruneSymNet: A Symbolic Neural Network and Pruning Algorithm for Symbolic Regression

Symbolic regression aims to derive interpretable symbolic expressions from data in order to better understand and interpret data. %which plays an important role in knowledge discovery and interpretable machine learning. In this study, a symbolic network called PruneSymNet is proposed for symbolic regression. This is a novel neural network whose activation function consists of common elementary functions and operators. The whole network is differentiable and can be trained by gradient descent method. Each subnetwork in the network corresponds to an expression, and our goal is to extract such subnetworks to get the desired symbolic expression. Therefore, a greedy pruning algorithm is proposed to prune the network into a subnetwork while ensuring the accuracy of data fitting. The proposed greedy pruning algorithm preserves the edge with the least loss in each pruning, but greedy algorithm often can not get the optimal solution. In order to alleviate this problem, we combine beam search during pruning to obtain multiple candidate expressions each time, and finally select the expression with the smallest loss as the final result. It was tested on the public data set and compared with the current popular algorithms. The results showed that the proposed algorithm had better accuracy.

cs.LG↗

arXiv2024

Linear subspaces of the intersection of two quadrics via Kuznetsov component

Let $Q_i(i=1,2)$ be $2g$ dimensional quadrics in $\mathbb{P}^{2g+1}$ and let $Y$ be the smooth intersection $Q_1\cap Q_2$. We associate the linear subspace in $Y$ with vector bundles on the hyperelliptic curve $C$ of genus $g$ by the left adjoint functor of $Φ:D^b(C)\rightarrow D^b(Y)$. As an application, we give a different proof of the classification of line bundles and stable bundles of rank $2$ on hyperelliptic curves given by Desale and Ramanan. When $g=3$, we show that the projection functor induces a closed embedding $α:Y\rightarrow SU^s_C(4,h)$ into the moduli space of stable bundles on $C$ of rank $4$ of fixed determinant.

math.AG↗

arXiv2023

RDE: A Hybrid Policy Framework for Multi-Agent Path Finding Problem

Multi-agent path finding (MAPF) is an abstract model for the navigation of multiple robots in warehouse automation, where multiple robots plan collision-free paths from the start to goal positions. Reinforcement learning (RL) has been employed to develop partially observable distributed MAPF policies that can be scaled to any number of agents. However, RL-based MAPF policies often get agents stuck in deadlock due to warehouse automation's dense and structured obstacles. This paper proposes a novel hybrid MAPF policy, RDE, based on switching among the RL-based MAPF policy, the Distance heat map (DHM)-based policy and the Escape policy. The RL-based policy is used for coordination among agents. In contrast, when no other agents are in the agent's field of view, it can get the next action by querying the DHM. The escape policy that randomly selects valid actions can help agents escape the deadlock. We conduct simulations on warehouse-like structured grid maps using state-of-the-art RL-based MAPF policies (DHC and DCC), which show that RDE can significantly improve their performance.

cs.RO↗

arXiv2023

Generating Transferable and Stealthy Adversarial Patch via Attention-guided Adversarial Inpainting

Adversarial patch attacks can fool the face recognition (FR) models via small patches. However, previous adversarial patch attacks often result in unnatural patterns that are easily noticeable. Generating transferable and stealthy adversarial patches that can efficiently deceive the black-box FR models while having good camouflage is challenging because of the huge stylistic difference between the source and target images. To generate transferable, natural-looking, and stealthy adversarial patches, we propose an innovative two-stage attack called Adv-Inpainting, which extracts style features and identity features from the attacker and target faces, respectively and then fills the patches with misleading and inconspicuous content guided by attention maps. In the first stage, we extract multi-scale style embeddings by a pyramid-like network and identity embeddings by a pretrained FR model and propose a novel Attention-guided Adaptive Instance Normalization layer (AAIN) to merge them via background-patch cross-attention maps. The proposed layer can adaptively fuse identity and style embeddings by fully exploiting priority contextual information. In the second stage, we design an Adversarial Patch Refinement Network (APR-Net) with a novel boundary variance loss, a spatial discounted reconstruction loss, and a perceptual loss to boost the stealthiness further. Experiments demonstrate that our attack can generate adversarial patches with improved visual quality, better stealthiness, and stronger transferability than state-of-the-art adversarial patch attacks and semantic attacks.

cs.CV↗

arXiv2023

A Survey of Robustness and Safety of 2D and 3D Deep Learning Models Against Adversarial Attacks

Benefiting from the rapid development of deep learning, 2D and 3D computer vision applications are deployed in many safe-critical systems, such as autopilot and identity authentication. However, deep learning models are not trustworthy enough because of their limited robustness against adversarial attacks. The physically realizable adversarial attacks further pose fatal threats to the application and human safety. Lots of papers have emerged to investigate the robustness and safety of deep learning models against adversarial attacks. To lead to trustworthy AI, we first construct a general threat model from different perspectives and then comprehensively review the latest progress of both 2D and 3D adversarial attacks. We extend the concept of adversarial examples beyond imperceptive perturbations and collate over 170 papers to give an overview of deep learning model robustness against various adversarial attacks. To the best of our knowledge, we are the first to systematically investigate adversarial attacks for 3D models, a flourishing field applied to many real-world applications. In addition, we examine physical adversarial attacks that lead to safety violations. Last but not least, we summarize present popular topics, give insights on challenges, and shed light on future research on trustworthy AI.

cs.LG↗

arXiv2022

Physical-World Optical Adversarial Attacks on 3D Face Recognition

2D face recognition has been proven insecure for physical adversarial attacks. However, few studies have investigated the possibility of attacking real-world 3D face recognition systems. 3D-printed attacks recently proposed cannot generate adversarial points in the air. In this paper, we attack 3D face recognition systems through elaborate optical noises. We took structured light 3D scanners as our attack target. End-to-end attack algorithms are designed to generate adversarial illumination for 3D faces through the inherent or an additional projector to produce adversarial points at arbitrary positions. Nevertheless, face reflectance is a complex procedure because the skin is translucent. To involve this projection-and-capture procedure in optimization loops, we model it by Lambertian rendering model and use SfSNet to estimate the albedo. Moreover, to improve the resistance to distance and angle changes while maintaining the perturbation unnoticeable, a 3D transform invariant loss and two kinds of sensitivity maps are introduced. Experiments are conducted in both simulated and physical worlds. We successfully attacked point-cloud-based and depth-image-based 3D face recognition algorithms while needing fewer perturbations than previous state-of-the-art physical-world 3D adversarial attacks.

cs.CV↗

arXiv2022

SimCC: a Simple Coordinate Classification Perspective for Human Pose Estimation

The 2D heatmap-based approaches have dominated Human Pose Estimation (HPE) for years due to high performance. However, the long-standing quantization error problem in the 2D heatmap-based methods leads to several well-known drawbacks: 1) The performance for the low-resolution inputs is limited; 2) To improve the feature map resolution for higher localization precision, multiple costly upsampling layers are required; 3) Extra post-processing is adopted to reduce the quantization error. To address these issues, we aim to explore a brand new scheme, called \textit{SimCC}, which reformulates HPE as two classification tasks for horizontal and vertical coordinates. The proposed SimCC uniformly divides each pixel into several bins, thus achieving \emph{sub-pixel} localization precision and low quantization error. Benefiting from that, SimCC can omit additional refinement post-processing and exclude upsampling layers under certain settings, resulting in a more simple and effective pipeline for HPE. Extensive experiments conducted over COCO, CrowdPose, and MPII datasets show that SimCC outperforms heatmap-based counterparts, especially in low-resolution settings by a large margin.

cs.CV↗

arXiv2021

Attend to Who You Are: Supervising Self-Attention for Keypoint Detection and Instance-Aware Association

This paper presents a new method to solve keypoint detection and instance association by using Transformer. For bottom-up multi-person pose estimation models, they need to detect keypoints and learn associative information between keypoints. We argue that these problems can be entirely solved by Transformer. Specifically, the self-attention in Transformer measures dependencies between any pair of locations, which can provide association information for keypoints grouping. However, the naive attention patterns are still not subjectively controlled, so there is no guarantee that the keypoints will always attend to the instances to which they belong. To address it we propose a novel approach of supervising self-attention for multi-person keypoint detection and instance association. By using instance masks to supervise self-attention to be instance-aware, we can assign the detected keypoints to their corresponding instances based on the pairwise attention scores, without using pre-defined offset vector fields or embedding like CNN-based bottom-up models. An additional benefit of our method is that the instance segmentation results of any number of people can be directly obtained from the supervised attention matrix, thereby simplifying the pixel assignment pipeline. The experiments on the COCO multi-person keypoint detection challenge and person instance segmentation task demonstrate the effectiveness and simplicity of the proposed method and show a promising way to control self-attention behavior for specific purposes.

cs.CV↗

arXiv2021

Generating Unrestricted 3D Adversarial Point Clouds

Utilizing 3D point cloud data has become an urgent need for the deployment of artificial intelligence in many areas like facial recognition and self-driving. However, deep learning for 3D point clouds is still vulnerable to adversarial attacks, e.g., iterative attacks, point transformation attacks, and generative attacks. These attacks need to restrict perturbations of adversarial examples within a strict bound, leading to the unrealistic adversarial 3D point clouds. In this paper, we propose an Adversarial Graph-Convolutional Generative Adversarial Network (AdvGCGAN) to generate visually realistic adversarial 3D point clouds from scratch. Specifically, we use a graph convolutional generator and a discriminator with an auxiliary classifier to generate realistic point clouds, which learn the latent distribution from the real 3D data. The unrestricted adversarial attack loss is incorporated in the special adversarial training of GAN, which enables the generator to generate the adversarial examples to spoof the target network. Compared with the existing state-of-art attack methods, the experiment results demonstrate the effectiveness of our unrestricted adversarial attack methods with a higher attack success rate and visual quality. Additionally, the proposed AdvGCGAN can achieve better performance against defense models and better transferability than existing attack methods with strong camouflage.

cs.CV↗

arXiv2021

TokenPose: Learning Keypoint Tokens for Human Pose Estimation

Human pose estimation deeply relies on visual clues and anatomical constraints between parts to locate keypoints. Most existing CNN-based methods do well in visual representation, however, lacking in the ability to explicitly learn the constraint relationships between keypoints. In this paper, we propose a novel approach based on Token representation for human Pose estimation~(TokenPose). In detail, each keypoint is explicitly embedded as a token to simultaneously learn constraint relationships and appearance cues from images. Extensive experiments show that the small and large TokenPose models are on par with state-of-the-art CNN-based counterparts while being more lightweight. Specifically, our TokenPose-S and TokenPose-L achieve $72.5$ AP and $75.8$ AP on COCO validation dataset respectively, with significant reduction in parameters ($\downarrow80.6\%$; $\downarrow$ $56.8\%$) and GFLOPs ($\downarrow$ $75.3\%$; $\downarrow$ $24.7\%$). Code is publicly available.

cs.CV↗

arXiv2021

Hidden Backdoor Attack against Semantic Segmentation Models

Deep neural networks (DNNs) are vulnerable to the \emph{backdoor attack}, which intends to embed hidden backdoors in DNNs by poisoning training data. The attacked model behaves normally on benign samples, whereas its prediction will be changed to a particular target label if hidden backdoors are activated. So far, backdoor research has mostly been conducted towards classification tasks. In this paper, we reveal that this threat could also happen in semantic segmentation, which may further endanger many mission-critical applications ($e.g.$, autonomous driving). Except for extending the existing attack paradigm to maliciously manipulate the segmentation models from the image-level, we propose a novel attack paradigm, the \emph{fine-grained attack}, where we treat the target label ($i.e.$, annotation) from the object-level instead of the image-level to achieve more sophisticated manipulation. In the annotation of poisoned samples generated by the fine-grained attack, only pixels of specific objects will be labeled with the attacker-specified target class while others are still with their ground-truth ones. Experiments show that the proposed methods can successfully attack semantic segmentation models by poisoning only a small proportion of training data. Our method not only provides a new perspective for designing novel attacks but also serves as a strong baseline for improving the robustness of semantic segmentation methods.

cs.CR↗

arXiv2019

Discrete Element Method Model of Elastic Fiber Uniaxial Compression

A flexible fiber model based on the discrete element method (DEM) is presented and validated for the simulation of uniaxial compression of flexible fibers in a cylindrical container. It is found that the contact force models in the DEM simulations have a significant impact on compressive forces exerted on the fiber bed. Only when the geometry-dependent normal contact force model and the static friction model are employed, the simulation results are in good agreement with experimental results. Systematic simulation studies show that the compressive force initially increases and eventually saturates with an increase in the fiber-fiber friction coefficient, and the fiber-fiber contact forces follow a similar trend. The compressive force and lateral shear-to-normal stress ratio increase linearly with increasing fiber-wall friction coefficient. In uniaxial compression of frictional fibers, more static friction contacts occur than dynamic friction contacts with static friction becoming more predominant as the fiber-fiber friction coefficient increases.

cond-mat.soft↗

arXiv2019

Yielding and hardening of flexible fiber packings during triaxial compression

This paper examines the mechanical response of flexible fiber packings subject to triaxial compression. Short fibers yield in a manner similar to typical granular materials in which the deviatoric stress remains nearly constant with increasing strain after reaching a peak value. Interestingly, long fibers exhibit a hardening behavior, where the stress increases rapidly with increasing strain at large strains and the packing density continuously increases. Phase diagrams for classifying the bulk mechanical response as yielding, hardening, or a transition regime are generated as a function of the fiber aspect ratio, fiber-fiber friction coefficient, and confining pressure. Large fiber aspect ratio, large fiber-fiber friction coefficient, and large confining pressure promote hardening behavior. The hardening packings can support much larger loads than the yielding packings contributing to the stability and consolidation of the granular structure, but larger internal axial forces occur within fibers.

cond-mat.soft↗

arXiv2018

Dynamic 3D shape measurement based on the phase-shifting moiré algorithm

In order to increase the efficiency of phase retrieval,Wang proposed a high-speed moire phase retrieval method.But it is used only to measure the tiny object. In view of the limitation of Wang method,we proposed a dynamic three-dimensional (3D) measurement based on the phase-shifting moire algorithm.First, four sinusoidal fringe patterns with a pi/2 phase-shift are projected on the reference plane and acquired four deformed fringe patterns of the reference plane in advance. Then only single-shot deformed fringe pattern of the tested object is captured in measurement process.Four moire fringe patterns can be obtained by numerical multiplication between the the AC component of the object pattern and the AC components of the reference patterns respectively. The four low-frequency components corresponding to the moire fringe patterns are calculated by the complex encoding FT (Fourier transform) ,spectrum filtering and inverse FT.Thus the wrapped phase of the object can be determined in the tangent form from the four phase-shifting moire fringe patterns using the four-step phase shifting algorithm.The continuous phase distribution can be obtained by the conventional unwrapping algorithm. Finally, experiments were conducted to prove the validity and feasibility of the proposed method. The results are analyzed and compared with those of Wang method, demonstrating that our method not only can expand the measurement scope, but also can improve accuracy.

physics.optics↗

arXiv2016

Improved phase-unwrapping method using geometric constraints

Conventional dual-frequency fringe projection algorithm often suffers from phase unwrapping failure when the frequency ratio between the high frequency and the low one is too large. Zhang et.al. proposed an enhanced two-frequency phase-shifting method to use geometric constraints of digital fringe projection(DFP) to reduce the noise impact due to the large frequency ratio. However, this method needs to calibrate the DFP system and calculate the minimum phase map at the nearest position from the camera perspective, these procedures are are relatively complex and more time-cosuming. In this paper, we proposed an improved method, which eliminates the system calibration and determination in Zhang's method,meanwhile does not need to use the low frequency fringe pattern. In the proposed method,we only need a set of high frequency fringe patterns to measure the object after the high frequency is directly estimated by the experiment. Thus the proposed method can simplify the procedure and improve the speed. Finally, the experimental evaluation is conducted to prove the validity of the proposed method.The results demonstrate that the proposed method can overcome the main disadvantages encountered by Zhang's method.

cs.CV↗

arXiv2016

One shot profilometry using iterative two-step temporal phase-unwrapping

This paper reviews two techniques that have been recently published for 3D profilometry and proposes one shot profilometry using iterative two-step temporal phase-unwrapping by combining the composite fringe projection and the iterative two-step temporal phase unwrapping algorithm. In temporal phase unwrapping, many images with different frequency fringe pattern are needed to project which would take much time. In order to solve this problem, Ochoa proposed a phase unwrapping algorithm based on phase partitions using a composite fringe, which only needs projecting one composite fringe pattern with four kinds of frequency information to complete the process of 3D profilometry. However, we found that the fringe order determined through the construction of phase partitions tended to be imprecise. Recently, we proposed an iterative two-step temporal phase unwrapping algorithm, which can achieve high sensitivity and high precision shape measurement. But it needs multiple frames of fringe images which would take much time. In order to take into account both the speed and accuracy of 3D shape measurement, we get a new, and more accurate unwrapping method based on composite fringe pattern by combining these two techniques. This method not only retains the speed advantage of Ochoa's algorithm, but also greatly improves its measurement accuracy. Finally, the experimental evaluation is conducted to prove the validity of the proposed method, and the experimental results show that this method is feasible.

physics.ins-det↗

arXiv2016

An algorithm to increase the residues of wrapped-phase in spatial domain

In phase unwrapping, the locations and densities of residues are indicative of the severity of the unwrapping problem. The residues are used to detect and evade inconsistent phase areas. Gdeisat et al. proposed an algorithm to increase the number of residues in a wrapped-phase map to improve the results of phase unwrapping. But this method will take much time to make the Fourier transform, inverse Fourier transform, select and shift the spectral components, and there is no theoretical analysis on why the frequency shift can increase the number of residues. In view of the above problems, we proposed an algorithm to increase the number of residues in a wrapped-phase map, which only uses a simple multiply operation in spatial domain to realize frequency shift by taking advantage of the frequency shift property of Fourier transform. Besides that, we discuss the relationship between the number of residues and frequency shift. Finally, the experimental evaluation is conducted to prove the validity of the proposed method. Experimental results demonstrated that the proposed method can speed up more than 50%.

physics.ins-det↗

arXiv2016

Improved method for phase wraps reduction in profilometry

In order to completely eliminate, or greatly reduce the number of phase wraps in 2D wrapped phase map, Gdeisat et al. proposed an algorithm, which uses shifting the spectrum towards the origin. But the spectrum can be shifted only by an integer number, meaning that the phase wraps reduction is often not optimal. In addition, Gdeisat's method will take much time to make the Fourier transform, inverse Fourier transform, select and shift the spectral components. In view of the above problems, we proposed an improved method for phase wraps elimination or reduction. First, the wrapped phase map is padded with zeros, the carrier frequency of the projected fringe is determined by high resolution, which can be used as the moving distance of the spectrum. And then realize frequency shift in spatial domain. So it not only can enable the spectrum to be shifted by a rational number when the carrier frequency is not an integer number, but also reduce the execution time. Finally, the experimental results demonstrated that the proposed method is feasible.

physics.ins-det↗