arXiv subjects

Pei Yang

Publications and source records attributed to Pei Yang.

At least 37 records · Page 2Linked to original sources

arXiv2025

IDProtector: An Adversarial Noise Encoder to Protect Against ID-Preserving Image Generation

Recently, zero-shot methods like InstantID have revolutionized identity-preserving generation. Unlike multi-image finetuning approaches such as DreamBooth, these zero-shot methods leverage powerful facial encoders to extract identity information from a single portrait photo, enabling efficient identity-preserving generation through a single inference pass. However, this convenience introduces new threats to the facial identity protection. This paper aims to safeguard portrait photos from unauthorized encoder-based customization. We introduce IDProtector, an adversarial noise encoder that applies imperceptible adversarial noise to portrait photos in a single forward pass. Our approach offers universal protection for portraits against multiple state-of-the-art encoder-based methods, including InstantID, IP-Adapter, and PhotoMaker, while ensuring robustness to common image transformations such as JPEG compression, resizing, and affine transformations. Experiments across diverse portrait datasets and generative models reveal that IDProtector generalizes effectively to unseen data and even closed-source proprietary models.

cs.CV↗

arXiv2025

Erlang Model for Multi-type Data Flow

With the development of information technology, requirements for data flow have become diverse. When multi-type data flow (MDF) is used, games, videos, calls, etc. are all requirements. There may be a constant switch between these requirements, and also multiple requirements at the same time. Therefore, the demands of users change over time, which makes traditional teletraffic analysis not directly applicable. This paper proposes probabilistic models for the requirement of MDF, and analyzes in three states: non-tolerance, tolerance and delay. When the requirement random variables are co-distributed with respect to time, we prove the practicability of the Erlang Multirate Loss Model (EMLM) from a mathematical perspective by discretizing time and error analysis. An algorithm of pre-allocating resources is given to guild the construction of base resources.

cs.NI↗

arXiv2024

Anti-Reference: Universal and Immediate Defense Against Reference-Based Generation

Diffusion models have revolutionized generative modeling with their exceptional ability to produce high-fidelity images. However, misuse of such potent tools can lead to the creation of fake news or disturbing content targeting individuals, resulting in significant social harm. In this paper, we introduce Anti-Reference, a novel method that protects images from the threats posed by reference-based generation techniques by adding imperceptible adversarial noise to the images. We propose a unified loss function that enables joint attacks on fine-tuning-based customization methods, non-fine-tuning customization methods, and human-centric driving methods. Based on this loss, we train a Adversarial Noise Encoder to predict the noise or directly optimize the noise using the PGD method. Our method shows certain transfer attack capabilities, effectively challenging both gray-box models and some commercial APIs. Extensive experiments validate the performance of Anti-Reference, establishing a new benchmark in image security.

cs.CV↗

arXiv2024

RingID: Rethinking Tree-Ring Watermarking for Enhanced Multi-Key Identification

We revisit Tree-Ring Watermarking, a recent diffusion model watermarking method that demonstrates great robustness to various attacks. We conduct an in-depth study on it and reveal that the distribution shift unintentionally introduced by the watermarking process, apart from watermark pattern matching, contributes to its exceptional robustness. Our investigation further exposes inherent flaws in its original design, particularly in its ability to identify multiple distinct keys, where distribution shift offers no assistance. Based on these findings and analysis, we present RingID for enhanced multi-key identification. It consists of a novel multi-channel heterogeneous watermarking approach designed to seamlessly amalgamate distinctive advantages from diverse watermarks. Coupled with a series of suggested enhancements, RingID exhibits substantial advancements in multi-key identification. Github Page: https://github.com/showlab/RingID

cs.CV↗

arXiv2024

Steganalysis on Digital Watermarking: Is Your Defense Truly Impervious?

Digital watermarking techniques are crucial for copyright protection and source identification of images, especially in the era of generative AI models. However, many existing watermarking methods, particularly content-agnostic approaches that embed fixed patterns regardless of image content, are vulnerable to steganalysis attacks that can extract and remove the watermark with minimal perceptual distortion. In this work, we categorize watermarking algorithms into content-adaptive and content-agnostic ones, and demonstrate how averaging a collection of watermarked images could reveal the underlying watermark pattern. We then leverage this extracted pattern for effective watermark removal under both graybox and blackbox settings, even when the collection contains multiple watermark patterns. For some algorithms like Tree-Ring watermarks, the extracted pattern can also forge convincing watermarks on clean images. Our quantitative and qualitative evaluations across twelve watermarking methods highlight the threat posed by steganalysis to content-agnostic watermarks and the importance of designing watermarking techniques resilient to such analytical attacks. We propose security guidelines calling for using content-adaptive watermarking strategies and performing security evaluation against steganalysis. We also suggest multi-key assignments as potential mitigations against steganalysis vulnerabilities.

cs.CV↗

arXiv2024

WMAdapter: Adding WaterMark Control to Latent Diffusion Models

Watermarking is crucial for protecting the copyright of AI-generated images. We propose WMAdapter, a diffusion model watermark plugin that takes user-specified watermark information and allows for seamless watermark imprinting during the diffusion generation process. WMAdapter is efficient and robust, with a strong emphasis on high generation quality. To achieve this, we make two key designs: (1) We develop a contextual adapter structure that is lightweight and enables effective knowledge transfer from heavily pretrained post-hoc watermarking models. (2) We introduce an extra finetuning step and design a hybrid finetuning strategy to further improve image quality and eliminate tiny artifacts. Empirical results demonstrate that WMAdapter offers strong flexibility, exceptional image generation quality and competitive watermark robustness.

cs.CV↗

arXiv2024

Modeling Hierarchical Structural Distance for Unsupervised Domain Adaptation

Unsupervised domain adaptation (UDA) aims to estimate a transferable model for unlabeled target domains by exploiting labeled source data. Optimal Transport (OT) based methods have recently been proven to be a promising solution for UDA with a solid theoretical foundation and competitive performance. However, most of these methods solely focus on domain-level OT alignment by leveraging the geometry of domains for domain-invariant features based on the global embeddings of images. However, global representations of images may destroy image structure, leading to the loss of local details that offer category-discriminative information. This study proposes an end-to-end Deep Hierarchical Optimal Transport method (DeepHOT), which aims to learn both domain-invariant and category-discriminative representations by mining hierarchical structural relations among domains. The main idea is to incorporate a domain-level OT and image-level OT into a unified OT framework, hierarchical optimal transport, to model the underlying geometry in both domain space and image space. In DeepHOT framework, an image-level OT serves as the ground distance metric for the domain-level OT, leading to the hierarchical structural distance. Compared with the ground distance of the conventional domain-level OT, the image-level OT captures structural associations among local regions of images that are beneficial to classification. In this way, DeepHOT, a unified OT framework, not only aligns domains by domain-level OT, but also enhances the discriminative power through image-level OT. Moreover, to overcome the limitation of high computational complexity, we propose a robust and efficient implementation of DeepHOT by approximating origin OT with sliced Wasserstein distance in image-level OT and accomplishing the mini-batch unbalanced domain-level OT.

cs.CV↗

arXiv2023

PMT-IQA: Progressive Multi-task Learning for Blind Image Quality Assessment

Blind image quality assessment (BIQA) remains challenging due to the diversity of distortion and image content variation, which complicate the distortion patterns crossing different scales and aggravate the difficulty of the regression problem for BIQA. However, existing BIQA methods often fail to consider multi-scale distortion patterns and image content, and little research has been done on learning strategies to make the regression model produce better performance. In this paper, we propose a simple yet effective Progressive Multi-Task Image Quality Assessment (PMT-IQA) model, which contains a multi-scale feature extraction module (MS) and a progressive multi-task learning module (PMT), to help the model learn complex distortion patterns and better optimize the regression issue to align with the law of human learning process from easy to hard. To verify the effectiveness of the proposed PMT-IQA model, we conduct experiments on four widely used public datasets, and the experimental results indicate that the performance of PMT-IQA is superior to the comparison approaches, and both MS and PMT modules improve the model's performance.

eess.IV↗

arXiv2023

TriangleNet: Edge Prior Augmented Network for Semantic Segmentation through Cross-Task Consistency

This paper addresses the task of semantic segmentation in computer vision, aiming to achieve precise pixel-wise classification. We investigate the joint training of models for semantic edge detection and semantic segmentation, which has shown promise. However, implicit cross-task consistency learning in multi-task networks is limited. To address this, we propose a novel "decoupled cross-task consistency loss" that explicitly enhances cross-task consistency. Our semantic segmentation network, TriangleNet, achieves a substantial 2.88\% improvement over the Baseline in mean Intersection over Union (mIoU) on the Cityscapes test set. Notably, TriangleNet operates at 77.4\% mIoU/46.2 FPS on Cityscapes, showcasing real-time inference capabilities at full resolution. With multi-scale inference, performance is further enhanced to 77.8\%. Furthermore, TriangleNet consistently outperforms the Baseline on the FloodNet dataset, demonstrating its robust generalization capabilities. The proposed method underscores the significance of multi-task learning and explicit cross-task consistency enhancement for advancing semantic segmentation and highlights the potential of multitasking in real-time semantic segmentation.

cs.CV↗

arXiv2021

Asynchronous Massive Access in Multi-cell Wireless Networks Using Reed-Muller Codes

Providing connectivity to a massive number of devices is a key challenge in 5G wireless systems. In particular, it is crucial to develop efficient methods for active device identification and message decoding in a multi-cell network with fading, path loss, and delay uncertainties. This paper presents such a scheme using second-order Reed-Muller (RM) sequences and orthogonal frequency-division multiplexing (OFDM). For given positive integer $m$, a codebook is generated with up to $2^{m(m+3)/2}$ codewords of length $2^m$, where each codeword is a unique RM sequence determined by a matrix-vector pair with binary entries. This allows every device to send $m(m + 3)/2$ bits of information where an arbitrary number of these bits can be used to represent the identity of a node, and the remaining bits represent a message. There can be up to $2^{m(m+3)/2}$ different identities. Using an iterative algorithm, an access point can estimate the matrix-vector pairs of each nearby device, as long as not too many devices transmit in the same frame. It is shown that both the computational complexity and the error performance of the proposed algorithm exceed another state-of-the-art algorithm. The device identification and message decoding scheme developed in this work can serve as the basis for grant-free massive access for billions of devices with hundreds of simultaneously active devices in each cell.

cs.IT↗

arXiv2021

Graph-based Visual-Semantic Entanglement Network for Zero-shot Image Recognition

Zero-shot learning uses semantic attributes to connect the search space of unseen objects. In recent years, although the deep convolutional network brings powerful visual modeling capabilities to the ZSL task, its visual features have severe pattern inertia and lack of representation of semantic relationships, which leads to severe bias and ambiguity. In response to this, we propose the Graph-based Visual-Semantic Entanglement Network to conduct graph modeling of visual features, which is mapped to semantic attributes by using a knowledge graph, it contains several novel designs: 1. it establishes a multi-path entangled network with the convolutional neural network (CNN) and the graph convolutional network (GCN), which input the visual features from CNN to GCN to model the implicit semantic relations, then GCN feedback the graph modeled information to CNN features; 2. it uses attribute word vectors as the target for the graph semantic modeling of GCN, which forms a self-consistent regression for graph modeling and supervise GCN to learn more personalized attribute relations; 3. it fuses and supplements the hierarchical visual-semantic features refined by graph modeling into visual embedding. Our method outperforms state-of-the-art approaches on multiple representative ZSL datasets: AwA2, CUB, and SUN by promoting the semantic linkage modelling of visual features.

cs.CV↗

arXiv2020

Exact dynamics of the homogeneous two-qubit $XXZ$ central spin model with the spin bath prepared in superpositions of symmetric Dicke states

We obtain exact dynamics of a two-qubit central spin model (CSM) consisting of two interacting qubits homogeneously coupled to a spin bath via the $XXZ$-type coupling, with the bath initially prepared in linear superpositions of the symmetric Dicke states. Using the interaction picture Hamiltonian with respect to the non-spin-flipping part of the model, we derive a sequence of equations of motion within each magnetization sector satisfied by the probability amplitudes of the time-evolved state. These equations of motion admit analytical solutions for the single-qubit CSM in which one of the two central qubits decouples from the rest of the system. Based on this, we provide a quantitative interpretation to the observed collapse-revival phenomena in the single-qubit Rabi oscillations when the bath is prepared in the spin coherent state. We then study the disentanglement and coherence dynamics of two initially entangled noninteracting qubits when the two qubits interact with individual baths or with a common bath. For individual baths the coherent dynamics is found to positively correlated to the single-qubit purity dynamics, and entanglement sudden disappearance and revivals are observed in both cases. The entanglement creation of two initially separable qubits coupled to a common bath is also studied and collapse and revival behaviors in the entanglement dynamics are observed. Choosing the equally weighted state and the $W$-class states as the bath initial states, we finally study the dynamics of entanglement between two individual bath spins and demonstrate the entanglement sharing mechanism in such a system.

quant-ph↗

arXiv2020

Massive Access in Multi-cell Wireless Networks Using Reed-Muller Codes

Providing connectivity to a massive number of devices is a key challenge in 5G wireless systems. In particular, it is crucial to develop efficient methods for active device identification and message decoding in a multi-cell network with fading and path loss uncertainties. In this paper, we design such a scheme using second-order Reed-Muller (RM) sequences. For given positive integer $m$, a codebook is generated with up to $2^{m(m+3)/2}$ codewords of length $2^m$, where each codeword is a unique RM sequence determined by a matrix-vector pair with binary entries. This allows every device to send $m(m+3)/2$ bits of information where an arbitrary number of these bits can be used to represent the identity of a node, and the remaining bits represent a message. There can be up to $2^{m(m+3)/2}$ devices in total. Using an iterative algorithm, an access point can estimate the matrix-vector pairs of each nearby device, as long as not too many devices transmit simultaneously. To improve the performance, we also describe an enhanced RM coding scheme with slotting. We show that both the computational complexity and the error performance of the latter algorithm exceed another state-of-the-art algorithm. The device identification and message decoding scheme developed in this work can serve as the basis for grant-free massive access for billions of devices with hundreds of simultaneously active devices in each cell.

cs.IT↗

arXiv2019

Parallelized Training of Restricted Boltzmann Machines using Markov-Chain Monte Carlo Methods

Restricted Boltzmann Machine (RBM) is a generative stochastic neural network that can be applied to collaborative filtering technique used by recommendation systems. Prediction accuracy of the RBM model is usually better than that of other models for recommendation systems. However, training the RBM model involves Markov-Chain Monte Carlo (MCMC) method, which is computationally expensive. In this paper, we have successfully applied distributed parallel training using Horovod framework to improve the training time of the RBM model. Our tests show that the distributed training approach of the RBM model has a good scaling efficiency. We also show that this approach effectively reduces the training time to little over 12 minutes on 64 CPU nodes compared to 5 hours on a single CPU node. This will make RBM models more practically applicable in recommendation systems.

cs.LG↗

arXiv2019

Ergodic Mutual Information for Generalized Fadings

In this paper, novel expressions are derived to evaluate the ergodic mutual information (EMI) under BPSK modulation of single-input single-output (SISO) systems operating in generalized fading channels, including $η$-$μ$ fading and $κ$-$μ$ fading. To verify our derivation, we first investigate the specific fading types, namely Rayleigh, Nakagami-$m$ and Rician, and then turn to generalized fading scenarios. It is shown that all the expressions concluded from the generalized cases can be specialized into those derived from the specific ones. Different from the conventional results of the EMI, our developed expressions contain only the simplest numerical calculations, without any Meijer's G-functions which must be implemented in the particular computing software. Additionally, it should be noted that our work provides a complete analysis of the EMI in wireless channel under discrete inputs.

eess.SP↗

arXiv2019

Performance of Multiple Antennas Selection in SIMO Systems with BPSK/QPSK Modulations

This paper studies the performance of single-input multiple-output (SIMO) systems under receive antenna selection (RAS) and BPSK/QPSK modulations. At the receiver, a subset of branches are selected and combined using maximal-ratio combining (MRC) to maximize the instantaneous Signal to Noise Ratio (SNR). By assuming independent and identical distributed (i.i.d.) Rayleigh flat fading, a closed-form expression, with considerably high precision, is developed to approximate the average input-output mutual information, also termed as symmetric capacity, of the whole system. Later, this approximated expression is further utilized to investigate the efficient capacity and energy efficiency of the SIMO system under BPSK/QPSK modulations and RAS. Besides analytical derivations, simulations are provided to demonstrate the approximation precision, feasibility and validity of the derived results.

eess.SP↗

arXiv2019

Ergodic Secrecy Rate of Antenna-Selection-Aided MIMOME Channels with BPSK/QPSK Modulations

This paper analyzes transmit antenna selection (TAS) under Rayleigh flat fading for BPSK/QPSK modulations in multiple-input multiple-output wiretap channels, also termed as multiple-input multiple-output multiple-eavesdropper (MIMOME) channels. In our protocol, a single antenna is selected to transmit the secret message and selection combing (SC) or maximal-ratio combing (MRC) is utilized at the legitimate receiver or the eavesdropper. Novel closed-form expressions for the ergodic secrecy rates are derived to approximate the exact values, which hold high precision and compact forms. Besides theoretical derivations, simulations are provided to demonstrate the feasibility and validity of the proposed formulas.

eess.SP↗

arXiv2019

Ergodic H-S/MRC Mutual Information

This paper studies the ergodic mutual information of hybrid selection/maximal-ratio combining (H-S/MRC) diversity system under BPSK/QPSK modulations. We consider a simple single-input multiple-output (SIMO) channel, where a subset of branches are selected and combined using maximal-ratio combining (MRC) to maximize the instantaneous Signal to Noise Ratio (SNR) at the receiver. For independent and identically distributed (i.i.d.) Rayleigh flat fading, a general recursive expression is developed to estimate the ergodic input-output mutual information of the whole system. Besides analytical derivations, simulations are provided to demonstrate the feasibility and validity of the derived results.

eess.SP↗