ManTra-Net: Manipulation Tracing Network for Detection and Localization of Image Forgeries With Anomalous Features

Wu, Yue; AbdAlmageed, Wael; Natarajan, Premkumar

doi:10.1109/cvpr.2019.00977

Cited by 376 publications

(384 citation statements)

References 42 publications

Supporting

Mentioning

382

Contrasting

Unclassified

Order By: Relevance

“…This is expected since this feature captures where else on the web this image appeared in the past, and out-of-context images are typically more likely to have appeared in low quality or fact checking websites, which makes them easily discernible. We do not observe any significant differences in accuracy for detecting manipulated images or memes, even using features specific to those images, whether features from a recent paper that detects potentially manipulated pixels (Wu et al 2019) or features about text in the image (see Methods).…”

Section: Image Type Popular Random Clusters Sharesmentioning

confidence: 76%

“…Identifying photoshopped images is hard, even manually (Nightingale et al 2017). State-of-the-art image processing techniques exist to detect manipulated JPEG images (Wu et al 2019). We tried using these techniques to identify if an image has been digitally manipulated, but these tools were not particularly useful here.…”

Section: Challenges In Detecting Image Misinformationmentioning

confidence: 99%

“…For instance, the out-ofcontext images are easy to detect using image hashing techniques, if they make use of images that appear in a database of content already shared in the past. Similarly, advances in computer vision techniques (Wu et al 2019) and image processing (Lee & Osindero 2016) to identify text in images can be used to detect manipulated images. Since each of these types make up a significant portion of the misinformation, even trying to identify a subset of such images automatically would be immensely helpful.…”

Section: Image Type Popular Random Clusters Sharesmentioning

confidence: 99%

“…We obtained the features related to objects, text and faces from the image. In addition to the content features from Cloud Vision, we also obtained the fraction of pixels in an image that have been manipulated (Wu et al 2019) and used a pre-trained convolutional neural network (CNN) trained with Facebook's ResNext architecture. We used a publicly available model 9 provided by PyTorch pre-trained on ImageNet.…”

Section: Machine Learning Modelsmentioning

confidence: 99%

See 3 more Smart Citations

Images and Misinformation in Political Groups: Evidence from WhatsApp in India

Garimella¹,

Eckles²

2020

HKS Misinfo Review

View full text Add to dashboard Cite

WhatsApp is a key medium for the spread of news and rumors, often shared as images. We study a large collection of politically-oriented WhatsApp groups in India, focusing on the period leading up to the 2019 Indian national elections. By labeling samples of random and popular images, we find that around 10% of shared images are known misinformation and most fall into three types of images. Machine learning methods can be used to predict whether a viral image is misinformation, but are brittle to shifts in content over time.

show abstract

Section: Image Type Popular Random Clusters Sharesmentioning

confidence: 76%

Section: Challenges In Detecting Image Misinformationmentioning

confidence: 99%

Section: Image Type Popular Random Clusters Sharesmentioning

confidence: 99%

Section: Machine Learning Modelsmentioning

confidence: 99%

See 2 more Smart Citations

Images and Misinformation in Political Groups: Evidence from WhatsApp in India

Garimella¹,

Eckles²

2020

HKS Misinfo Review

View full text Add to dashboard Cite

show abstract

“…For the first category, we selected the SPAM+SVM method [16], winner of the First IEEE Forensic Challenge and based on the SPAM steganalytic features [23], the CNN+SVM method of [24], which extract features through a constrained CNN, LSTM-EnDec [32], which uses a long-short term memory recurrent neural network to detect pristine/forged spatial transitions, and MantraNet [44], which performs joint image-level detection and pixel-level localization of forgeries, regarded as local image anomalies. For the second category, we consider several forgery localization methods converted into image-level detectors.…”

Section: ) Reference Methodsmentioning

confidence: 99%

A Full-Image Full-Resolution End-to-End-Trainable CNN Framework for Image Forgery Detection

et al. 2020

View full text Add to dashboard Cite

Due to limited computational and memory resources, current deep learning models accept only rather small images in input, calling for preliminary image resizing. This is not a problem for high-level vision problems, where discriminative features are barely affected by resizing. On the contrary, in image forensics, resizing tends to destroy precious high-frequency details, impacting heavily on performance. One can avoid resizing by means of patch-wise processing, at the cost of renouncing whole-image analysis. In this work, we propose a CNN-based image forgery detection framework which makes decisions based on full-resolution information gathered from the whole image. Thanks to gradient checkpointing, the framework is trainable end-to-end with limited memory resources and weak (image-level) supervision, allowing for the joint optimization of all parameters. Experiments on widespread image forensics datasets prove the good performance of the proposed approach, which largely outperforms all baselines and all reference methods. INDEX TERMS CNN, Digital image forensics, forgery detection.

show abstract

Dual‐Stream Fusion and Multi‐scale Analysis: Introducing the Synergistic Dual‐Stream Network (SDS‐Net) for Image Manipulation Segmentation

Song,

Lin,

Xie

2024

Advanced Intelligent Systems

View full text Add to dashboard Cite

This article introduces the synergistic dual‐stream network (SDS‐Net), a novel neural network architecture that significantly enhances the detection of image manipulations. SDS‐Net employs a unique dual‐stream fusion strategy that processes both RGB image and the corresponding noise map. It innovatively combines features computation from the different blocks of dual backbones, and leverages a multi‐scale spatial pyramid pooling (MSPP) module to expand the receptive fields of shallow features. This approach not only enriches the feature representation, but also ensures the precise localization of manipulated regions. Extensive experiments conducted on various public datasets demonstrate the superiority of SDS‐Net over several state‐of‐the‐art methods.

show abstract

ManTra-Net: Manipulation Tracing Network for Detection and Localization of Image Forgeries With Anomalous Features

Cited by 376 publications

References 42 publications

Images and Misinformation in Political Groups: Evidence from WhatsApp in India

Images and Misinformation in Political Groups: Evidence from WhatsApp in India

A Full-Image Full-Resolution End-to-End-Trainable CNN Framework for Image Forgery Detection

Dual‐Stream Fusion and Multi‐scale Analysis: Introducing the Synergistic Dual‐Stream Network (SDS‐Net) for Image Manipulation Segmentation

Contact Info

Product

Resources

About