OSCAR-Net: Object-centric Scene Graph Attention for Image Attribution

  title={OSCAR-Net: Object-centric Scene Graph Attention for Image Attribution},
  author={Eric Nguyen and Tu Bui and Vishy Swaminathan and John P. Collomosse},
  journal={2021 IEEE/CVF International Conference on Computer Vision (ICCV)},
Images tell powerful stories but cannot always be trusted. Matching images back to trusted sources (attribution) enables users to make a more informed judgment of the images they encounter online. We propose a robust image hashing algorithm to perform such matching. Our hash is sensitive to manipulation of subtle, salient visual details that can substantially change the story told by an image. Yet the hash is invariant to benign transformations (changes in quality, codecs, sizes, shapes, etc… 

Figures and Tables from this paper

RepMix: Representation Mixing for Robust Attribution of Synthesized Images

RepMix 4, the GAN fingerprinting technique based on representation mixing and a novel loss, is proposed and validated, showing its capability of tracing the provenance of GAN-generated images invariant to the semantic content of the image and also robust to perturbations.

A Self-Supervised Descriptor for Image Copy Detection

This work introduces SSCD, a model that builds on a recent self-supervised contrastive training objective, and adapts this method to the copy detection task, including a pooling operator from the instance matching literature, and adapting contrastive learning to augmentations that combine images.

VPN: Video Provenance Network for Robust Content Attribution

This work uses an inverted index to match temporal chunks of video using late-fusion to combine both visual and audio features, and demonstrates high accuracy recall over a corpus of 100,000 videos.

A survey of visual neural networks: current trends, challenges and opportunities

The basic architectures and current trends of these two types of VNNs, Convolutional neural networks and Transformer networks are introduced and the lightweight, robust and interpretable solutions are summarized and analyzed.

ARIA: Adversarially Robust Image Attribution for Content Provenance

An approach to prevent imperceptible adversarial attacks on deep visual fingerprinting models, via robust contrastive learning, and how to train an adversarially robust image comparator model for detecting editorial changes in matched images.



SceneSketcher: Fine-Grained Image Retrieval with Scene Sketches

This paper proposes a graph embedding based method to learn the similarity measurement between images and scene sketches, which models the multi-modal information, including the size and appearance of objects as well as their layout information, in an effective manner.

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

The forgery localization problem is formulated as a local anomaly detection problem, a Z-score feature is designed to capture local anomaly, and a novel long short-term memory solution is proposed to assess local anomalies.

Specifying Object Attributes and Relations in Interactive Scene Generation

  • Oron AshualLior Wolf
  • Computer Science
    2019 IEEE/CVF International Conference on Computer Vision (ICCV)
  • 2019
The method separates between a layout embedding and an appearance embedding, which leads to generated images that better match the scene graph, have higher visual quality, and support more complex scene graphs.

Multi-Label Image Recognition With Graph Convolutional Networks

This work proposes a multi-label classification model based on Graph Convolutional Network (GCN), and proposes a novel re-weighted scheme to create an effective label correlation matrix to guide information propagation among the nodes in GCN.

CNN-Generated Images Are Surprisingly Easy to Spot… for Now

It is demonstrated that, with careful pre- and post-processing and data augmentation, a standard image classifier trained on only one specific CNN generator (ProGAN) is able to generalize surprisingly well to unseen architectures, datasets, and training methods.

Image Provenance Analysis at Scale

An end-to-end processing pipeline for image provenance analysis which works at real-world scale is presented, employing a cutting-edge image filtering solution that is custom-tailored for the problem at hand, as well as novel techniques for obtaining the provenance graph that expresses how the images, as nodes, are ancestrally connected.

Aligning Linguistic Words and Visual Semantic Units for Image Captioning

This work proposes to explicitly model the object interactions in semantics and geometry based on Graph Convolutional Networks (GCNs), and fully exploit the alignment between linguistic words and visual semantic units for image captioning.

The PS-Battles Dataset - an Image Collection for Image Manipulation Detection

The PS-Battles dataset is presented, which is gathered from a large community of image manipulation enthusiasts and provides a basis for media derivation and manipulation detection in the visual domain.

Discovering Image Manipulation History by Pairwise Relation and Forensics Tools

This paper proposes to learn a pairwise ancestor-offspring detector and use it jointly with a holistic image manipulation detector to identify the source image of the small regions that are spliced into the probe image and, detecting source images within the search results.

Deep Hashing Network for Efficient Similarity Retrieval

A novel Deep Hashing Network (DHN) architecture for supervised hashing is proposed, in which good image representation tailored to hash coding and formally control the quantization error are jointly learned.