50,054 Hits in 3.2 sec

Learning Transformation-Aware Embeddings for Image Forensics [article]

Aparna Bharati, Daniel Moreira, Patrick Flynn, Anderson Rocha, Kevin Bowyer, Walter Scheirer
2020 arXiv   pre-print
Our approach learns transformation-aware descriptors using weak supervision via composited transformations and a rank-based quadruplet loss.  ...  One of the main sub-problems for provenance analysis that has not yet been addressed directly is the edit ordering of images that share full content or are near-duplicates.  ...  We categorize them broadly to explain how advances in techniques for these general tasks become useful in designing a pipeline to learn transformation-aware embeddings for ordering images in provenance  ... 
arXiv:2001.04547v1 fatcat:4vz2jdn62bfcze5w3qagdpytra

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

Eric Nguyen, Tu Bui, Vishy Swaminathan, John Collomosse
2021 arXiv   pre-print
Our key contribution is OSCAR-Net (Object-centric Scene Graph Attention for Image Attribution Network); a robust image hashing model inspired by recent successes of Transformers in the visual domain.  ...  The network is trained via contrastive learning on a dataset of original and manipulated images yielding a state of the art image hash for content fingerprinting that scales to millions of images.  ...  Acknowledgement We thank Andy Parsons, Leonard Rosenthol, Bill Marino and the Adobe Content Authenticity Initiative (CAI) for discussions.  ... 
arXiv:2108.03541v2 fatcat:vzs5ciwi5jhuzlpwpqh7wfx2wu

FakeTagger: Robust Safeguards against DeepFake Dissemination via Provenance Tracking [article]

Run Wang, Felix Juefei-Xu, Meng Luo, Yang Liu, Lina Wang
2021 arXiv   pre-print
to recover the embedded message after various drastic GAN-based DeepFake transformation with high confidence.  ...  The embedded message could be employed to represent the identity of facial images, which further contributed to DeepFake detection and provenance.  ...  [60] proposed to apply a novel transformation-aware adversarially perturbed faces to disrupt the DeepFake creation.  ... 
arXiv:2009.09869v3 fatcat:eemjt2fnxrgr7eg4dursbec5hu

LAMPRET: Layout-Aware Multimodal PreTraining for Document Understanding [article]

Te-Lin Wu, Cheng Li, Mingyang Zhang, Tao Chen, Spurthi Amba Hombaiah, Michael Bendersky
2021 arXiv   pre-print
To bridge this gap, we parse a document into content blocks (eg. text, table, image) and propose a novel layout-aware multimodal hierarchical framework, LAMPreT, to model the blocks and the whole document  ...  We evaluate the proposed model on two layout-aware tasks -- text block filling and image suggestion and show the effectiveness of our proposed hierarchical architecture as well as pretraining techniques  ...  A pretrained CNN module is adopted for all the models and baselines to encode the images, and then transformed to the same embedding size of the Wordpiece token embedding, 768, with an MLP layer.  ... 
arXiv:2104.08405v1 fatcat:k2ss7rzs5ngqlgsds44ybgvf5i

A Novel Data Analytics Oriented Approach for Image Representation Learning in Manufacturing Systems

Yue Liu, Junqi Ma, Xingzhen Tao, Jingyun Liao, Tao Wang, Jingjing Chen, Haidong Shao
2022 Journal of Sensors  
In this paper, we propose a novel self-supervised self-attention learning framework—TriLFrame for image representation learning.  ...  The TriLFrame is based on the hybrid architecture of Convolutional Network and Transformer.  ...  Although it is proven that positional embeddings make self-attention operation be aware of sequential information to some degree [44, 45] , when using self-attention on image patches, the goal is to embed  ... 
doi:10.1155/2022/1807103 fatcat:rlnobzugkfgnpdd6acxgsbz5ma

Vehicle Re-identification Based on Quadratic Split Architecture and Auxiliary Information Embedding

Tongwei LU, Hao ZHANG, Feng MIN, Shihai JIA
2022 IEICE Transactions on Fundamentals of Electronics Communications and Computer Sciences  
(II) The Auxiliary Information Embedding (AIE) is proposed to improve the robustness of the model by plugging a learnable camera/viewpoint embedding into Vit.  ...  More precisely, we split an image into many patches as "global part" and further split them into smaller sub-patches as "local part".  ...  [10] has proven that pure transformer can achieve excellent feature extraction results in image recognition tasks just like CNN.  ... 
doi:10.1587/transfun.2022eal2008 fatcat:hf7uyck3zngzbl6lw743yhxlcy

The effectiveness of discrete hermite wavelet filters technique in digital image watermarking

Areej M. Abduldaim, Asma Abdulelah Abdulrahman, Fouad Shaker Tahir
2022 Indonesian Journal of Electrical Engineering and Computer Science  
<span>In this work, new discrete wavelets were derived Hermite polynomials for obtained discrete hermite wavelet transformation (DHWT), and their efficiency for use in image processing is demonstrated  ...  The color image has been subjected to various attacks after which the watermark is retrieved from the image after comparing it with the proposed algorithm and it has proven its power faster and better  ...  Watermark with video using lefting wavelet transform (LWT) transformations for the gray watermark to be more transmission using information jamming by embedding with video attacks [10] , [15] - [18]  ... 
doi:10.11591/ijeecs.v25.i3.pp1392-1399 fatcat:v67hn5q6u5c27hnb3x4aggjpui

Contrastive Semantic Similarity Learning for Image Captioning Evaluation with Intrinsic Auto-encoder [article]

Chao Zeng, Tiesong Zhao, Sam Kwong
2021 arXiv   pre-print
Motivated by the auto-encoder mechanism and contrastive representation learning advances, we propose a learning-based metric for image captioning, which we call Intrinsic Image Captioning Evaluation(I^  ...  Automatically evaluating the quality of image captions can be very challenging since human language is quite flexible that there can be various expressions for the same meaning.  ...  Triplet loss was proven useful in FaceNet [32] to learn distance aware face embedding, which means similar images would be pushed together and images with contrast differences located in distant area  ... 
arXiv:2106.15312v1 fatcat:r54o6bp4grhw3gcaihdw377msy

Rethinking Reprojection: Closing the Loop for Pose-aware ShapeReconstruction from a Single Image [article]

Rui Zhu, Hamed Kiani Galoogahi, Chaoyang Wang, Simon Lucey
2017 arXiv   pre-print
Our evaluation on several object categories demonstrates the superiority of our method for predicting pose-aware 3D shapes from natural images.  ...  In this paper we define the new task of pose-aware shape reconstruction from a single image, and we advocate that cheaper 2D annotations of objects silhouettes in natural images can be utilized.  ...  Conclusion We define the new task of pose-aware shape reconstruction from a single natural image, and update the recent methods of TL-embedding Network and 3D-VAE-GAN to close the loop for this task, in  ... 
arXiv:1707.04682v2 fatcat:x4l4h7i7jnevzk3lpackhx3yqi

Embed Everything: A Method for Efficiently Co-Embedding Multi-Modal Spaces [article]

Sarah Di, Robin Yu, Amol Kapoor
2021 arXiv   pre-print
We prove the use of this system in a joint image-audio embedding task.  ...  In this paper, we propose a novel and cost-effective HTL strategy for co-embedding multi-modal spaces.  ...  As far as we are aware, HTL methods have not been applied to learning joint embedding spaces in multiple modalities, such as images and audio.  ... 
arXiv:2110.04599v1 fatcat:xtkozzcxmrdnbgrx3qompyjmai

Visually Aware Skip-Gram for Image Based Recommendations [article]

Parth Tiwari, Yash Jain, Shivansh Mundra, Jenny Harding, Manoj Kumar Tiwari
2020 arXiv   pre-print
We propose a novel framework VASG (Visually Aware Skip-Gram) for learning user and product representations in a common latent space using product image features.  ...  image features to the Skip-Gram embedding space.  ...  Images have also been proven to be useful for Point-of-Interest recommendations in [34] and for tag recommendations in [27] .  ... 
arXiv:2008.06908v1 fatcat:64l4dps7z5fcbh7snpuztxyc4y

Boundary-aware Information Maximization for Self-supervised Medical Image Segmentation [article]

Jizong Peng, Ping Wang, Marco Pedersoli, Christian Desrosiers
2022 arXiv   pre-print
However, it is not trivial to build reasonable pairs for a segmentation task in an unsupervised way.  ...  Unsupervised pre-training has been proven as an effective approach to boost various downstream tasks given limited labeled data.  ...  For dense embedding, positive pairs are embeddings in the same position undergoing different intensity transformations, where negative pairs are defined as embeddings with sufficient large distances.  ... 
arXiv:2202.02371v2 fatcat:sup2l4iahjdl3dajxcytmmmwsq

Mixture of gaussian models and bayes error under differential privacy

Bowei Xi, Murat Kantarcioglu, Ali Inan
2011 Proceedings of the first ACM conference on Data and application security and privacy - CODASPY '11  
Utility Aware Redaction • Since rules could be applied in any order, heuristics need for ordering • We choose three conventions for pre-ordering the production rules: -the original ordering (OO); -lowest  ...  rule application to proceed, -the attribute element describes the annotations in LHS. • Embedding element has two optional sub elements, -pre describes how LH S is connected to the provenance graph -post  ... 
doi:10.1145/1943513.1943537 dblp:conf/codaspy/XiKI11 fatcat:ly7lnvp6ejan3bbt56ilhm5xgm

Security issues on digital watermarking algorithms

Wioletta Wójtowicz, Marek R. Ogiela
2012 Annales UMCS Informatica  
In the paper we provide an overview of the most popular digital watermarking methods for still images available today.  ...  Popular transforms of images include the DFT (Discrete Fourier Transform) ([1, 2, 3, 4, 5]), DCT (Discrete Cosine Transform) ([1, 3, 6, 5]) and DWT (Discrete Wavelet Transform) ([1, 3, 4, 7, 6, 5]).  ...  for embedding the watermark in the host data.  ... 
doi:10.2478/v10065-012-0021-3 fatcat:tmsiasxchfdzpbrwatgc2el76i

Toward Exploiting Second-Order Feature Statistics for Arbitrary Image Style Transfer

Hyun-Chul Choi
2022 Sensors  
First, a new correlation-aware loss and a correlation-aware feature alignment technique are proposed.  ...  However, the previous approaches used feature alignment techniques that were too simple in their transform layer to cover the characteristics of style features of images.  ...  The network was trained to minimize the sum of content and style losses for a specific style embedding.  ... 
doi:10.3390/s22072611 pmid:35408228 pmcid:PMC9003536 fatcat:v4lkjtpo3ba5vg6zdgigkw3vty
« Previous Showing results 1 — 15 out of 50,054 results