只把伪影相关特征送入分类器

发布时间：2025-06-24 17:16:52  作者：北方职教升学中心  阅读量：751


前言：本篇博客总结 2018-2023年 通用AIGI（AI-Generated Image）检测相关研究工作。
核心思想：CLIP+MLP，不同的是，生成数据是由真实数据修改得到，修改方法包括频域掩膜/patch swapping/图像融合/颜色迁移，以此增强检测结果的泛化性。抹除图像的中频部分，重建得到伪重建图像，与原始图像的重建图像concate进行分类

  • A Bias-Free Training Paradigm for More General AI-generated Image Detection. arXiv, 20241223.
    Fabrizio Guillaro, Giada Zingarini, Ben Usman, Avneesh Sud, Davide Cozzolino, Luisa Verdoliva.
    核心思想：

    • PS：如有遗漏，欢迎评论区补充～

  • Fusing Global and Local Features for Generalized AI-Synthesized Image Detection. ICIP 2022
    Yan Ju, Shan Jia, Lipeng Ke, Hongfei Xue, Koki Nagano, Siwei Lyu.
    核心思想：使用ResNet，提全局特征和local patches的特征，融合之后分类检测

    • 2023

    • GLFF: Global and Local Feature Fusion for AI-synthesized Image Detection. IEEE Transactions on Multimedia, 2023.
      Yan Ju, Shan Jia, Jialing Cai, Haiying Guan, Siwei Lyu.
      核心思想：ICIP 2022 那篇的扩充版本

    • Learning on Gradients: Generalized Artifacts Representation for GAN-Generated Images Detection. CVPR, 2023.
      Chuangchuang Tan, Yao Zhao, Shikui Wei, Guanghua Gu, Yunchao Wei.
      核心思想：利用梯度信息，检测GAN生成的图像

    • Towards Universal Fake Image Detectors that Generalize Across Generative Models. CVPR, 2023.
      Utkarsh Ojha, Yuheng Li, Yong Jae Lee.
      核心思想：使用CLIP提特征，然后基于此特征完成检测（1）分类器；（2）基于特征之间的距离

    • De-fake: Detection and attribution of fake images generated by text-to-image generation models. CCS, 2023.
      Zeyang Sha, Zheng Li, Ning Yu, Yang Zhang.
      核心思想：主要利用提示词与对应图像的距离，设计简单分类器完成检测。

    • FingerprintNet: Synthesized Fingerprints for Generated Image Detection. ECCV, 2022.
      Yonghyun Jeong, Doyeon Kim, Youngmin Ro, Pyounggeon Kim, and Jongwon Choi.
      核心思想：使用real生成synthesized图像，然后用于训练分类器

    • Discovering Transferable Forensic Features for CNN-Generated Images Detection. ECCV 2022.
      Singapore University of Technology and Design
      核心思想：利用图像的颜色信息，进行针对性地数据增强

    • BiHPF: bilateral high pass filters for robust deepfake detection. WACV, 2022.
      Jeong, Y., Kim, D., Min, S., Joe, S., Gwon, Y., Choi, J.
      核心思想：使用对抗训练的方式得到图像的伪影压缩图，证实artifact的存在，后使用双边高通滤波（像素域&频域）放大频域伪影送入分类器检测。使用颜色统计数据进行检测

    • Whodunit: Detection and Attribution of Synthetic Images by Leveraging Model-specific Fingerprints
      Alexander Wißmann, Steffen Zeiler, Robert M. Nickel, Dorothea Kolossa. MAD Workshop, 2024.
      核心思想：使用DCT、

    • Diffusion Noise Feature: Accurate and Fast Generated Image Detection. arXiv 2023.
      Yichi Zhang, Xiaogang Xu.
      核心思想：通过寻找diffusion模型生成图像中特有的噪声模式进行检测。基于SSIM比较的检测

    • Organic or Diffused: Can We Distinguish Human Art from AI-generated Images?
      Anna Yoo Jeong Ha, Josephine Passananti, Ronik Bhaskar, Shawn Shan, Reid Southen, Haitao Zheng, Ben Y. Zhao. CCS, 2024. 芝加哥大学（郑海涛）
      核心思想：使用现成的在线工具或DIRE/DEFAKE判别艺术品是人工绘制还是模型生成

    • Deep Image Fingerprint: Towards Low Budget Synthetic Image Detection and Model Lineage Analysis
      Sergey Sinitsa, Ohad Fried. WACV, 2024
      核心思想：training free，基于频域特征与两类参考图像的相似性. 测试集SD/Glide/MJ/DALLE

    • CLIPping the Deception: Adapting Vision-Language Models for Universal Deepfake Detection
      Sohail Ahmed Khan, Duc-Tien Dang-Nguyen. ICMR, 2024.
      核心思想：基于CLIP的Prompt tuning，Adapter，Fine-tuning，Linear Probing四种方式

    • FIDAVL: Fake Image Detection and Attribution using Vision-Language Model. ICPR, 2024.
      Mamadou Keita, Wassim Hamidouche, Hessen Bougueffa Eutamene, Abdelmalik Taleb-Ahmed, Abdenour Hadid.
      核心思想：基于Q-Former和Vicuna 7b

    • Harnessing the Power of Large Vision Language Models for Synthetic Image Detection
      Mamadou Keita, Wassim Hamidouche, Hassen Bougueffa, Abdenour Hadid, Abdelmalik Taleb-Ahmed. arXiv, 2024.
      核心思想：基于 BLIP 2 和 Q-Former 的分类器

    • Bi-LORA: A Vision-Language Approach for Synthetic Image Detection
      Mamadou Keita, Wassim Hamidouche, Hessen Bougueffa Eutamene, Abdenour Hadid, Abdelmalik Taleb-Ahmed. arXiv, 2024.
      核心思想：同样是基于BLIP的检测

    • SemGIR: Semantic-Guided Image Regeneration Based Method for AI-generated Image Detection and Attribution. ACM Multimedia, 2024.
      Xiao Yu, Kejiang Chen, Kai Zeng, Han Fang, Zijin Yang, Xiuwei Shang, Yuang Qi, Weiming Zhang, Nenghai Yu
      核心思想：few-shot场景，基于基于语义重建的参考图像生成，训练检测/溯源分类器

    • Unsupervised Generative Fake Image Detector. TCSVT, 2024.
      Tong Qiao, Hang Shao, Shichuang Xie, and Ran Shi.
      核心思想：无监督学习

    • CSC-Net: Cross-Color Spatial Co-Occurrence Matrix Network for Detecting Synthesized Fake Images
      Tong Qiao, Yuxing Chen, Xiaofei Zhou, Ran Shi, Hang Shao, Kunye Shen, and Xiangyang Luo. TCDS, 2024.
      核心思想：使用Cross-color空间共现矩阵，利用颜色信息进行检测。

    • Forensic analysis of AI-compression traces in spatial and frequency domain
      Sandra Bergmann, Denise Moussa, Fabian Brand, André Kaup, Christian Riess. PRL, 2024.
      核心思想：像素域和频域特征的结合

    • A guided-based approach for deepfake detection: RGB-depth integration via features fusion
      Giorgio Leporoni, Luca Maiano, Lorenzo Papa, Irene Amerini. PRL, 2024.
      核心思想：backbone提原始图像与深度图双路特征，接attention机制特征融合，后送入二分类器

    • Enhancing the Generalization of Synthetic Image Detection Models through the Exploration of Features in Deep Detection Models
      Alireza Hajabdollah Javaheri, Hossein Motamednia, Ahmad Mahmoudi-Azanveh. MVIP, 2024.
      核心思想：使用DIRE得到重建误差，将其放到CNN分类器中进行检测

    • Fake-GPT: Detecting Fake Image via Large Language Model. PRCV, 2024.
      Yuming Fan, Dongming Yang, Jiguang Zhang, Bang Yang, and Yuexian Zou
      核心思想：将图像表示成RGB数值序列作为输入，微调大语言模型Qwen，跨模态检测。相较于生成图像，真实图像对细微的扰动更加鲁棒，使用DINOv2提取扰动前和扰动后图像的特征

    • Development of a Dual-Input Neural Model for Detecting AI-Generated Imagery
      Jonathan Gallagher, William Pugsley. arXiv 20240619 University of Waterloo
      核心思想：使用像素域和频域的双路网络构建分类器

    • Improving Interpretability and Robustness for the Detection of AI-Generated Images
      Tatiana Gaintseva, Laida Kushnareva, German Magai, Irina Piontkovskaya, Sergey Nikolenko, Marting Benning,Serguei Barannikov, Gregory Slabaugh. arXiv, 20240621. UK/Russia/Japan/France
      核心思想：

    • Towards More Accurate Fake Detection on Images Generated from Advanced Generative and Neural Rendering Models. arXiv, 20241113.
      Chengdong Dong, Vijayakumar Bhagavatula, Zhenyu Zhou, Ajay Kumar. CMU/港城
      核心思想：数据集包括Sora和神经渲染模型的图像

    • Understanding and Improving Training-Free AI-Generated Image Detections with Vision Foundation Models. arXiv, 20241128.
      Chung-Ting Tsai, Ching-Yun Ko, I-Hsin Chung, Yu-Chiang Frank Wang, Pin-Yu Chen IBM/国立台湾大学
      核心思想：改进RIGID，提出ContrastiveBlur更好检测facial image；MINDER降低noise bias

    • Any-Resolution AI-Generated Image Detection by Spectral Learning. arXiv, 20241128.
      Dimitrios Karageorgiou, Symeon Papadopoulos, Ioannis Kompatsiaris, Efstratios Gavves CERTH
      核心思想：光谱学习 spectral learning

    • FIRE: Robust Detection of Diffusion-Generated Images via Frequency-Guided Reconstruction Error. arXiv, 20241210.
      Beilin Chu, Xuan Xu, Xin Wang, Yufei Zhang, Weike You, Linna Zhou. 北邮
      核心思想：DIRE的衍生工作。

      –
      AIGI检测相关博客如下：

      • 论文研读｜AI生成图像检测发展历程及研究现状
      • 针对AIGC检测的鲁棒性测试——常见攻击手段汇总
      • 论文研读｜以真实图像为参考依据的AIGC检测
      • 论文研读｜针对文生图模型的AIGC检测

      目录

      • 2018
      • 2019
      • 2020
      • 2021
      • 2022
      • 2023
      • 2024

      2018

      • Detection of GAN-generated Fake Images over Social Networks. MIPR, 2018.
        Francesco Marra, Diego Gragnaniello, Davide Cozzolino, Luisa Verdoliva.
        核心思想：GAN discriminator，Steganalysis features，CNN networks

      2019

      • Do GANs Leave Artificial Fingerprints? MIPR, 2019.
        Francesco Marra, Diego Gragnaniello, Luisa Verdoliva, Giovanni Poggi.
        核心思想：基于GAN噪声残差相似度的检测

      • Detecting and Simulating Artifacts in GAN Fake Images. WIFS, 2019.
        Xu Zhang, Svebor Karaman, and Shih-Fu Chang.
        核心思想：基于GAN上采样产生的的频域独特伪影

      • Incremental learning for the detection and classification of GAN-generated images. WIFS, 2019.
        Francesco Marra, Cristiano Saltori†, Giulia Boato and Luisa Verdoliva
        核心思想：增量学习

      2020

      • CNN-generated images are surprisingly easy to spot…for now. CVPR, 2020.
        Sheng-Yu Wang, Oliver Wang, Richard Zhang, Andrew Owens, Alexei A. Efros.
        核心思想：将数据增强添加至检测分类器，全面测试泛化性。虚假重建这4类图像，使用对比学习损失训练分类器，得到更加准确的决策边界。

      • Exposing the Fake: Effective Diffusion-Generated Images Detection (SeDID)
        Ruipeng Ma, Jinhao Duan, Fei Kong, Xiaoshuang Shi, Kaidi Xu. ICML Workshops, 2024. 电子科大
        核心思想：借鉴成员推断攻击的方法，利用重建生成的中间过程完成检测

      • Breaking Semantic Artifacts for Generalized AI-generated Image Detection. NeurIPS, 2024.
        Chende Zheng, Chenhao Lin, Zhengyu Zhao, Hang Wang, Xu Guo, Shuai Liu, Chao Shen
        核心思想：cross-scene, patch shuffle（排除语义缺陷过拟合的影响） & patch based feature extractor + patch feature flatten

      • Frequency-Aware Deepfake Detection: Improving Generalizability through Frequency Space Domain Learning. AAAI, 2024. (FreqNet)
        Chuangchuang Tan, Yao Zhao, Shikui Wei, Guanghua Gu, Ping Liu, Yunchao Wei
        核心思想：仅使用图像的高频特征完成检测，并针对性修改分类器，使其关注高频特征部分

      • Leveraging Representations from Intermediate Encoder-Blocks for Synthetic Image Detection. ECCV, 2024. (RINE)
        Christos Koutlis, Symeon Papadopoulos. CERTH
        核心思想：使用编码器中间层的fine-grained特征，训练一个重要性评估模块，交叉熵损失+对比损失

      • Contrasting Deepfakes Diffusion via Contrastive Learning and Global-Local Similarities. ECCV 2024.
        Lorenzo Baraldi, Federico Cocchi, Marcella Cornia, Lorenzo Baraldi, Alessandro Nicolosi, Rita Cucchiara. Italy
        核心思想：以往的CLIP只关注到了全局特征的提取，本文使用对比学习(InfoNCE loss)对齐全局和局部区域的相似性，添加数据增强模块模拟图像处理操作。

      • Leveraging Frequency Analysis for Deep Fake Image Recognition. ICML, 2020.
        Joel Frank, Thorsten Eisenhofer, Lea Schönherr, Asja Fischer, Dorothea Kolossa, Thorsten Holz.
        核心思想：基于DFT频域分析，同上篇文章

      • T-GD: Transferable GAN-generated Images Detection Framework. ICML 2020.
        Hyeonseong Jeon, Youngoh Bang, Junyaup Kim, Simon S. Woo.
        核心思想：使用教师——学生模型，self-training半监督学习

      • On the use of Benford’s law to detect GAN-generated images
        N. Bonettini, P. Bestagini, S. Milani, and S. Tubaro. ICPR, 2020.
        核心思想：使用本福特定律相关特征+随机森林

      • GAN-Generated Image Detection With Self-Attention Mechanism Against GAN Generator Defect. IEEE Journal of Selected Topics in Signal Processing, 2020.
        Zhongjie Mi, Xinghao Jiang, Tanfeng Sun, and Ke Xu.
        核心思想：自注意力机制

      2021

      • Beyond the Spectrum: Detecting Deepfakes via Re-Synthesis. IJCAI, 2021.
        Yang He, Ning Yu, Margret Keuper, Mario Fritz
        核心思想：使用去噪等图像再生成的方式得到残差，使用二分类器完成检测

      • A closer look at fourier spectrum discrepancies for cnn-generated images detection. CVPR, 2021.
        Keshigeyan Chandrasegaran, Ngoc-Trung Tra, Ngai-Man Cheung. SUTD
        核心思想：频谱差异&最后一个上采样操作

      • Are GAN generated images easy to detect? A critical analysis of the state-of-the-art. ICME, 2021.
        D. Gragnaniello, D. Cozzolino, F. Marra, G. Poggi and L. Verdoliva.
        核心思想：总结性文章

      • Detection of GAN-Generated Images by Estimating Artifact Similarity. SPL, 2021.
        Weichuang Li, Peisong He, et al.
        核心思想：对于生成图像和真实图像分别构建参考特征，然后分别比较测试图像特征与二个参考特征，根据相似度分数进行检测

      2022

      • Detecting Generated Images by Real Images. ECCV, 2022.
        Bo Liu, Fan Yang, Xiuli Bi, Bin Xiao, Weisheng Li, Xinbo Gao.
        核心思想：利用真实图像特有的噪声模式，设计简单分类器完成检测。功率谱密度(Power Spectral Density, PSD)（频域）以及自相关系数（像素域）训练分类器。

      • Detecting Artificial Intelligence-Generated images via deep trace representations and interactive feature fusion. Image Fusion, 2024.
        Qiang Xu, Xinghao Jiang, Tanfeng Sun, Hao Wang, Laijin Meng, Hong Yan.
        核心思想：

      • CIFAKE: Image Classification and Explainable Identification of AI-Generated Synthetic Images
        Jordan J. Bird, Ahmad Lotfi. IEEE Access, 2024.
        核心思想：CNN分类，Grad-CAM对检测结果解释

      • Enhancing Interpretability in AI-Generated Image Detection with Genetic Programming
        Mingqian Lin, Lin Shang, Xiaoying Gao. ICDM Workshop, 2024.
        核心思想：使用遗传编程提高检测方法的可解释性

      • On the Exploitation of DCT-Traces in the Generative-AI Domain
        Orazio Pontorno, Luca Guarnera, and Sebastiano Battiato. ICIP, 2024.
        核心思想：使用DCT变换进行检测

      • MDTL-NET: Computer-generated image detection based on multi-scale deep texture learning
        Qiang Xu, Shan Jia, Xinghao Jiang, Tanfeng Sun, Zhe Wang, and Hong Yan. Expert Systems with Applications, 2024.
        核心思想：设计多种网络堆叠，注意力机制等区分电脑生成的图像和真实图像

      • Did You Note My Palette? Unveiling Synthetic Images Through Color Statistics
        Lea Uhlenbrock, Davide Cozzolino, Denise Moussa, Luisa Verdoliva, and Christian Riess. IH&MMSec, 2024.
        核心思想：本文指出Perceptual loss 的使用使得两类图像在亮度上比色度上差异更大。

      • Text Modality Oriented Image Feature Extraction for Detecting Diffusion-based DeepFake
        Di Yang, Yihao Huang, Qing Guo, Felix Juefei-Xu, Xiaojun Jia, Run Wang, Geguang Pu, and Yang Liu. arXiv 20240528 华东师范大学&新加坡南洋理工&纽约大学&武汉大学
        核心思想：方法简称：TOFE，本工作只针对text-to-image模型，首先得到target image的text embedding，利用图像的high-level和low-level特征作为检测依据。

      • Improving Synthetically Generated Image Detection in Cross-Concept Settings. MAD 2023.
        Pantelis Dogoulis, Giorgos Kordopatis-Zilos, Ioannis Kompatsiaris, Symeon Papadopoulos
        核心思想：使用图像质量评价，选择生成图像训练集的子集，与真实图像一起送入ResNet

      • Depth map guided triplet network for deepfake face detection. Neural Networks, 2023.
        Buyun Liang, Zhongyuan Wang, Baojin Huang, Qin Zou, Qian Wang, Jingjing Lian.
        核心思想：使用triplet loss约束生成RGB图像特征和深度图特征，concat送入二分类器

      • AI-Generated Image Detection using a Cross-Attention Enhanced Dual-Stream Network. APSIPA ASC, 2023.
        Ziyi Xi, Wenmin Huang, Kangkang Wei, Weiqi Luo and Peijia Zheng.
        核心思想：基于交叉注意力的双流检测网络；residual stream + content stream; cross-attention; feature fusion

      • Generalizable Synthetic Image Detection via Language-guided Contrastive Learning. arXiv, 2023.
        Haiwei Wu, Jiantao Zhou, and Shile Zhang.
        核心思想：设计图像文本对，基于对比学习损失微调CLIP的image encoder和text encoder，使得虚假图像特征更接近“fake photo”提示文本特征。解码重建图像一同送入ResNet进行分类

      • LaRE^2: Latent Reconstruction Error Based Method for Diffusion-Generated Image Detection
        Yunpeng Luo, Junlong Du, Ke Yan, Shouhong Ding. CVPR, 2024.
        核心思想：对DIRE的改进，基于重建损失，提高检测效率

      • AEROBLADE: Training-Free Detection of Latent Diffusion Images Using Autoencoder Reconstruction Error
        Jonas Ricker, Denis Lukovnikov, and Asja Fischer. CVPR, 2024.
        核心思想：也是基于重建损失，自编码器，使用lpips值作为衡量指标

      • Raising the Bar of AI-generated Image Detection with CLIP
        Davide Cozzolino, Giovanni Poggi, Riccardo Corvi, Matthias Nießner, Luisa Verdoliva. CVPR Workshop, 2024.
        核心思想：使用CLIP提特征，然后使用简单的SVM对特征进行分类。

      • AntifakePrompt: Prompt-Tuned Vision-Language Models are Fake Image Detectors. arXiv 2023.
        You-Ming Chang, Chen Yeh, Wei-Chen Chiu, Ning Yu.
        核心思想：针对指定的instruction“Is this photo real S*”，训练LLM tokenizer和Q-former tokenizer对S*的表示，并获得检测结果。

      • Online Detection of AI-Generated Images. ICCV Workshop, 2023.
        David C. Epstein, Ishan Jain, Oliver Wang, Richard Zhang.
        核心思想：不仅探究了AIGC检测方法的泛化性，而且还测试了基于AI的图像篡改检测与定位的性能，实现了像素级的AIGC检测。真实重建、

      • DIRE for Diffusion-Generated Image Detection. ICCV, 2023.
        Zhendong Wang, Jianmin Bao, Wengang Zhou, Weilun Wang, Hezhen Hu, Hong Chen, Houqiang Li.
        核心思想：利用真实图像与合成图像重建前后的残差差异

      • On The Detection of Synthetic Images Generated by Diffusion Models. ICASSP, 2023
        Riccardo Corvi, Davide Cozzolino, Giada Zingarini, Giovanni Poggi, Koki Nagano, Luisa Verdoliva…
        核心思想：意大利LV团队做的，基于频域分析

      • Exposing fake images generated by text-to-image diffusion models. PRL, 2023.
        Qiang Xu, Hao Wang, Laijin Meng, Zhongjie Mi, Jianye Yuan, Hong Yan.
        核心思想：基于注意力机制的特征提取和基于ViT的特征提取，baselines主要选取的是针对自然图像和电脑生成图像的检测方法。

      • Fourier Spectrum Discrepancies in Deep Network Generated Images. NeurIPS, 2020.
        Tarik Dzanic, Karan Shah, Freddie Witherden.
        核心思想：基于DFT的高频信号，在高分辨率/低压缩率的情况下频谱特征更加容易区分。

      • Detecting Images Generated by Deep Diffusion Models Using Their Local Intrinsic Dimensionality. ICCV Workshop, 2023.
        Peter Lorenz, Ricard L. Durall, Janis Keuper.
        核心思想：利用对抗样本检测中常用的Local Intrinsic Dimensionality手段，检测AIGC图像。

      • Artifact Feature Purification for Cross-domain Detection of AI-generated Images
        Zheling Meng, Bo Peng, Jing Dong, Tieniu Tan, Haonan Cheng. Computer Vision and Image Understanding, 2024.
        核心思想：DFT频域&可学习的正交分解空域（特征显式纯化，把图像特征分解为伪影相关特征和伪影不相关特征）+互信息估计器增大两种特征之间的距离（特征隐式纯化）。虚假、解码噪声、

      • Watch your Up-Convolution: CNN Based Generative Deep Neural Networks are Failing to Reproduce Spectral Distributions. CVPR, 2020.
        Ricard Durall, Margret Keuper, Janis Keuper.
        核心思想：up-convolution操作无法复制真实图像频谱，基于此发现设计简单的检测方法。
        · Faster Than Lies: Real-time Deepfake Detection using Binary Neural Networks
        Romeo Lanzino, Federico Fontana, Anxhelo Diko, Marco Raoul Marini, Luigi Cinque. CVPR Workshops, 2024.
        核心思想：使用FFT频域特征和Local Binary Pattern纹理特征以及像素域三路构建分类器

      • MaskSim: Detection of synthetic images by masked spectrum similarity analysis.
        Li et al. CVPR Workshops, 2024. France/Brazil/Hongkong
        核心思想：

      • DistilDIRE: A Small, Fast, Cheap and Lightweight Diffusion Synthesized Deepfake Detection
        Yewon Lim, Changyeon Lee, Aerin Kim, Oren Etzioni. ICML, 2024.
        核心思想：对DIRE的改进，提高检测效率

      • How to Trace Latent Generative Model Generated Images without Artificial Watermark?
        Zhenting Wang, Vikash Sehwag, Chen Chen, Lingjuan Lyu, Dimitris N. Metaxas, Shiqing Ma. ICML, 2024.
        核心思想：使得重建误差最小的inverted latent

      • DRCT: Diffusion Reconstruction Contrastive Training towards Universal Detection of Diffusion Generated Images ICML 2024
        Baoying Chen et al. 阿里巴巴 & 中山大学
        核心思想：首先得到真实图像和虚假图像各自的重建图像，然后基于真实、

      • Detecting Generated Images by Real Images Only. arXiv 2023.
        Xiuli Bi, Bo Liu, Fan Yang, Bin Xiao, Weisheng Li, Gao Huang, Pamela C. Cosman.
        核心思想：把AIGC检测看成新颖点检测任务，训练一个基于正样本的单分类器。

      • RIGID: A Training-free and Model-Agnostic Framework for Robust AI-Generated Image Detection
        Zhiyuan He, Pin-Yu Chen and Tsung-Yi Ho. arXiv, 20240530. 港中文
        核心思想：training-free，基于相似度分数的检测。

      • X-Transfer: A Transfer Learning-Based Framework for Robust GAN-Generated Fake Image Detection. arXiv 2023.
        Lei Zhang, Hao Chen, Shu Hu, Bin Zhu, Xi Wu, Jinrong Hu, Xin Wang.
        核心思想：一个主网络，一个辅助网络，三路损失，加权反向传播，训练分类器。

      • Intriguing properties of synthetic images: from generative adversarial networks to diffusion models. CVPR Workshop, 2023.
        Riccardo Corvi, Davide Cozzolino, Giovanni Poggi, Koki Nagano, Luisa Verdoliva.
        核心思想：意大利LV团队做的，基于频域分析

      • Synthbuster: Towards Detection of Diffusion Model Generated Images. IEEE Open Journal of Signal Processing, 2023.
        Quentin Bammey.
        核心思想：高通滤波得到图像残差，然后经傅立叶变换得到频谱图，送入分类器检测。只把伪影相关特征送入分类器。此外，提出Diffusion-generated Deepfake Detection dataset (D3 )数据集，包括三个版本SD+DeepFloyd IF模型生成的图像

      • Zero-Shot Detection of AI-Generated Images. ECCV, 2024. (ZED)
        Davide Cozzolino, Giovanni Poggi, Matthias Nieÿner, and Luisa Verdoliva
        核心思想：基于coding cost的zero-shot检测，借鉴文本生成中对下一个token的预测

      • Evolution of Detection Performance throughout the Online Lifespan of Synthetic Images. ECCV Workshops, 2024.
        Dimitrios Karageorgiou, Quentin Bammey, Valentin Por cellini, Bertrand Goupil, Denis Teyssou, and Symeon Pa padopoulos.
        核心思想：

      • ZeroFake: Zero-Shot Detection of Fake Images Generated and Edited by Text-to-Image Generation Models. CCS 2024.
        Zeyang Sha, Yicong Tan, Mingjie Li, Michael Backes, Yang Zhang
        核心思想：真实图像和生成图像对于「对抗提示」的重建程度不同。

      • Tiny Autoencoders are Effective Few-Shot Generative Model Detectors
        Luca Bindini, Giulia Bertazzini, Daniele Baracchi, Dasara Shullani, Paolo Frasconi, and Alessandro Piva. WIFS, 2024. Italy
        核心思想：

      • Frequency Masking for Universal Deepfake Detection
        Chandler Timm Doloriel, Ngai-Man Cheung. ICASSP 2024. SUTD
        核心思想：提出一种数据增强策略，像素域random masking/patch masking+频域masking

      • Mastering Deepfake Detection: A Cutting-edge Approach to Distinguish GAN and Diffusion-model Images. TOMM. 2024
        Luca Guarnera, Oliver Giudice, Sebastiano Battiato. Italy
        核心思想：层级多标签分类——real/synthestic➡️GAN/DM➡️which model

      • Parents and Children: Distinguishing Multimodal Deepfakes from Natural Images. TOMM. 2024.
        Roberto Amoroso, Davide Morelli, Marcella Cornia, Lorenzo Baraldi, Alberto Del Bimbo, Rita Cucchiara.
        核心思想：1-real+N-fake；解藕语义信息和风格信息，提出COCOFake数据集

      • Detecting Computer-Generated Images by Using Only Real Images
        Ji Li, Kai Wang. ICMV, 2024. EI检索。

      • GenDet: Towards Good Generalizations for AI-Generated Image Detection. arXiv 2023.
        Mingjian Zhu, Hanting Chen, Mouxiao Huang, Wei Li, Hailin Hu, Jie Hu, Yunhe Wang.
        核心思想：基于教师学生模型的异常检测

      2024

      • Rethinking the Up-Sampling Operations in CNN-based Generative Network for Generalizable Deepfake Detection
        Chuangchuang Tan, Huan Liu, Yao Zhao, Shikui Wei, Guanghua Gu, Ping Liu, Yunchao Wei. CVPR, 2024.
        核心思想：基于上采样的痕迹特征，Neighboring Pixel Relationships（NPR）

      • Shadows Don’t Lie and Lines Can’t Bend! Generative Models don’t know Projective Geometry…for now
        Ayush Sarkar, Hanlin Mai, Amitabh Mahapatra, Svetlana Lazebnik, D.A. Forsyth, Anand Bhattad. CVPR, 2024.
        核心思想：使用影射几何学的知识，比如影子，透视场和线段

      • Forgery-aware Adaptive Transformer for Generalizable Synthetic Image Detection
        Huan Liu, Zichang Tan, Chuangchuang Tan, Yunchao Wei, Yao Zhao, Jingdong Wang. CVPR, 2024.
        核心思想：image encoder 和 text encoder 结合的分类器

      • FakeInversion: Learning to Detect Images from Unseen Text-to-Image Models by Inverting Stable Diffusion
        George Cazenavette, Avneesh Sud, Thomas Leung Ben Usman. CVPR, 2024.
        核心思想：原始图像、

      • Disentangling Different Levels of GAN Fingerprints for Task-specific Forensics. Computer Standards & Interfaces, 2023.
        Chi Liu, Tianqing Zhu, Yuan Zhao, Jun Zhang, Wanlei Zhou.
        核心思想：分别提取空域和频域的特征，用于不同类型的取证任务。

      • Rich and Poor Texture Contrast: A Simple yet Effective Approach for AI-generated Image Detection
        Nan Zhong, Yiran Xu, Zhenxing Qian, Xinpeng Zhang. arXiv, 20240307. 复旦大学
        核心思想：对于真实图像和AIGC图像的平滑块和纹理块之间的对比差异不同，设计分类器进行检测。

      • Data-Independent Operator: A Training-Free Artifact Representation Extractor for Generalizable Deepfake Detection
        Chuangchuang Tan, Ping Liu, RenShuai Tao, Huan Liu, Yao Zhao, Baoyuan Wu, Yunchao Wei. arXiv, 20240311. 北交
        核心思想：

      • Deepfake Detection without Deepfakes: Generalization via Synthetic Frequency Patterns Injection
        Davide Alessandro Coccomini, Roberto Caldelli, Claudio Gennaro, Giuseppe Fiameni, Giuseppe Amato, Fabrizio Falchi. arXiv, 20240320. Italy
        核心思想：

      • Fake or JPEG? Revealing Common Biases in Generated Image Detection Datasets
        Patrick Grommelt, Louis Weiss, Franz-Josef Pfreundt, Janis Keuper. arXiv, 20240328. Fraunhofer ITWM
        核心思想：

      • D^3 Scaling Up Deepfake Detection by Learning from Discrepancy
        Yongqi Yang, Zhihao Qian, Ye Zhu, Yu Wu. arXiv, 20240406. 武汉大学
        核心思想：

      • Mixture of Low-rank Experts for Transferable AI-Generated Image Detection
        Zihan Liu, Hanyi Wang, Yaoyu Kang, Shilin Wang. arXiv, 20240407. 上交
        核心思想：

      • Detecting Image Attribution for Text-to-Image Diffusion Models in RGB and Beyond
        Katherine Xu, Lingzhi Zhang, Jianbo Shi. arXiv, 20240410. 宾夕法尼亚大学
        核心思想：

      • A Single Simple Patch is All You Need for AI-generated Image Detection
        Jiaxuan Chen, Jieteng Yao, and Li Niu. arXiv, 20240420. 南航&上交
        核心思想：对测试图像随机裁剪，选择最简单的patch进行resize之后送到SRM Conv中，接ResNet50完成检测，方法简称：SSP

      • Let Real lmages be as a Judger, Spotting Fake lmages Synthesized with GenerativeModels
        Ziyou Liang, Run Wang, Weifeng Liu, Yuyang Zhang, Wenyuan Yang, Lina Wang, Xingkai Wang. arXiv, 20240525. 武大&中山大学
        核心思想：反复使用对比学习，提取出真实图像的同质特征，与虚假图像的特征拼接，送入检测器。

    上一篇：播放这个视频需要一个新的编解码器是什么意思？解决方案来了
    下一篇：基于springboottt基于spring vue uniapp语言课学习系统小程序