分类 - IML
IMDL-BenCo:A Comprehensive Benchmark and Codebase for Image Manipulation Detection & Localization
IMDL-BenCo:A Comprehensive Benchmark and Codebase for Image Manipulation Detection & Localization

发表于NeurIPS 2024,因为图像篡改检测没有统一的标准,所以构建一个全面的基准,并且设计了一个框架将部分sota网络集成:Mantra-Net,MVSS-net,CAT-Net,ObjectFormer,PSCC-Net,NCL-IML,Trufor和IML-ViT。

1
DiffForensics:Leveraging Diffusion Prior to Image Forgery Detection and Localization
DiffForensics:Leveraging Diffusion Prior to Image Forgery Detection and Localization

发表于CVPR2024,两阶段的训练过程,该框架包括自监督去噪扩散的训练前阶段和多任务微调阶段,提出了一种新的边缘提示增强模块,该模块集成在多个尺度的解码器中,以增强被篡改的边缘痕迹从粗到细。

2
SUMI-IFL:An Information-Theoretic Framework for Image Forgery Localization with Sufficiency and Minimality Constraints
SUMI-IFL:An Information-Theoretic Framework for Image Forgery Localization with Sufficiency and Minimality Constraints

发表于aixiv,使用信息瓶颈理论完成图像篡改任务,没和NP++、IFL-VIT比较。

3
EditGuard:Versatile Image Watermarking for Tamper Localization and Copyright Protection
EditGuard:Versatile Image Watermarking for Tamper Localization and Copyright Protection

发表于CVPR2024,将版权水印和图像篡改主动保护两个任务联合起来。

4
UGEE-Net:Uncertainty-guided and edge-enhanced network for image splicing localization
UGEE-Net:Uncertainty-guided and edge-enhanced network for image splicing localization

发表于NeuralNetworks 2024。

5
Deep Fuzzy K-Means With Adaptive Loss and Entropy Regularization
Deep Fuzzy K-Means With Adaptive Loss and Entropy Regularization

发表于IEEE Transactions on Fuzzy Systems 2019,提出了深度模糊k-means(DFKM),具有加权自适应损失函数的FKM。

6
Robust deep fuzzy K-means clustering for image data
Robust deep fuzzy K-means clustering for image data

发表于JCR一区、CCF B类期刊的Pattern Recognition 2024,提出了一种新的边界引导图像篡改定位模型,该模型通过精心设计的注意力和对比学习机制充分利用被篡改区域的边界信息,利用拉普拉斯正则化方法对隶属度矩阵进行约束,使从相似样本中学习到的隶属度也相互关联,将自适应损失函数引入到统一的框架中,可以减少各种异常值的影响,有助于增强聚类的鲁棒性。

7
AdaIFL:Adaptive Image Forgery Localization via a Dynamic and Importance-aware Transformer Network
AdaIFL:Adaptive Image Forgery Localization via a Dynamic and Importance-aware Transformer Network

发表于ECCV2024,提出了AdaIFL,为不同的网络组件定制不同的专家组,构建多个不同的特征子空间,利用自适应激活的专家网络,AdaIFL可以捕获与伪造模式相关的判别特征,增强了模型的泛化能力。提出了一种特征重要性感知注意力,自适应地感知不同区域的重要性,并将区域特征聚集成可变长度的标记,将模型的注意力导向更有区别和信息的区域。

8
Image Manipulation Detection With Implicit Neural Representation and Limited Supervision
Image Manipulation Detection With Implicit Neural Representation and Limited Supervision

发表于ECCV 2024。

9
Unified Frequency-Assisted Transformer Framework for Detecting and Grounding Multi-modal Manipulation
Unified Frequency-Assisted Transformer Framework for Detecting and Grounding Multi-modal Manipulation

发表于IJCV 2024,多模态的图像篡改检测。

10