Filters








6,383 Hits in 5.3 sec

Decomposition-Based Domain Adaptation for Real-World Font Recognition [article]

Zhangyang Wang, Jianchao Yang, Hailin Jin, Eli Shechtman, Aseem Agarwala, Jonathan Brandt, Thomas S. Huang
2015 arXiv   pre-print
We present a domain adaption framework to address a domain mismatch between synthetic training and real-world testing data.  ...  This real-to-synthetic domain gap caused poor generalization to new real data in previous font recognition methods (Chen et al. (2014)).  ...  synthetic and real-world data, and a second high-level sub-network that learns a deep classifier using the low-level features.  ... 
arXiv:1412.5758v4 fatcat:lsen6fgutvctxhar3lsbwxhnxy

DeepFont

Zhangyang Wang, Jianchao Yang, Hailin Jin, Eli Shechtman, Aseem Agarwala, Jonathan Brandt, Thomas S. Huang
2015 Proceedings of the 23rd ACM international conference on Multimedia - MM '15  
Next, to combat the domain mismatch between available training and testing data, we introduce a Convolutional Neural Network (CNN) decomposition approach, using a domain adaptation technique based on a  ...  We study the Visual Font Recognition (VFR) problem [4] , and advance the state-of-the-art remarkably by developing the DeepFont system.  ...  real-world data, and a high-level sub-network that learns a deep classifier from the low-level features.  ... 
doi:10.1145/2733373.2806219 dblp:conf/mm/WangYJSABH15 fatcat:3aio35s3cvatjmmwli2tlh3z7e

HENet: Forcing a Network to Think More for Font Recognition [article]

Jingchao Chen, Shiyi Mu, Shugong Xu, Youdong Ding
2021 arXiv   pre-print
The pluggable module hides the most discriminative accessible features and forces the network to consider other complicated features to solve the hard examples of similar fonts, called HE Block.  ...  This paper proposes a novel font recognizer with a pluggable module solving the font recognition task.  ...  However, the real-world test set has an extremely great difficulty due to the domain shift. We use AdobeVFR to evaluate the font recognizing performance on word images.  ... 
arXiv:2110.10872v1 fatcat:hnadhv27avg65ju74yvbmzergu

Economic Efficiency of Innovative Projects of CNN Modified Architecture Application

Viktor Khavalko, Andriana Mazur, Vladyslav Mykhailyshyn, Roman Zhelizniak, Iryna Kovtyk
2019 International Workshop on Cyber Hygiene  
The paper deals with involves the use of a modified architecture of a convolutional neural network to solve the problem of recognizing the Cyrillic alphabet letters in real time and with high accuracy.  ...  The analysis of the existing approaches and methods of handwriting recognition is carried out, features and the basic difficulties which arise at the decision of a recognition problem are considered.  ...  The development of deep neural networks [1] for image recognition contributes to the development of already known research areas in machine learning. One such area is domain adaptation (DA) [2] .  ... 
dblp:conf/cybhyg/KhavalkoMMZK19 fatcat:r4byi6ukrvbozkncnn5ye5lgxy

Studying Very Low Resolution Recognition Using Deep Networks [article]

Zhangyang Wang, Shiyu Chang, Yingzhen Yang, Ding Liu, Thomas S. Huang
2016 arXiv   pre-print
Taking advantage of techniques primarily in super resolution, domain adaptation and robust regression, we formulate a dedicated deep learning method and demonstrate how these techniques are incorporated  ...  The resulting Robust Partially Coupled Networks achieves feature enhancement and recognition simultaneously.  ...  VLRR Font Recognition Dataset The VFR dataset used in [6, 34, 33] includes 2,383 font classes.  ... 
arXiv:1601.04153v2 fatcat:5zubparjhjdghegis2drngicva

Domain Generalization by Solving Jigsaw Puzzles [article]

Fabio Maria Carlucci, Antonio D'Innocente, Silvia Bucci, Barbara Caputo, Tatiana Tommasi
2019 arXiv   pre-print
Multiple experiments on the PACS, VLCS, Office-Home and digits datasets confirm our intuition and show that this simple method outperforms previous domain generalization and adaptation solutions.  ...  In this paper we propose to apply a similar approach to the task of object recognition across domains: our model learns the semantic labels in a supervised fashion, and broadens its understanding of the  ...  Office-Home Art Clipart Product Real-World Avg.  ... 
arXiv:1903.06864v2 fatcat:acscjv6uqrdzvcl67xmy4vrxii

Studying Very Low Resolution Recognition Using Deep Networks

Zhangyang Wang, Shiyu Chang, Yingzhen Yang, Ding Liu, Thomas S. Huang
2016 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)  
Taking advantage of techniques primarily in super resolution, domain adaptation and robust regression, we formulate a dedicated deep learning method and demonstrate how these techniques are incorporated  ...  The resulting Robust Partially Coupled Networks achieves feature enhancement and recognition simultaneously.  ...  VLRR Font Recognition Dataset The VFR dataset used in [6, 34, 33] includes 2,383 font classes.  ... 
doi:10.1109/cvpr.2016.518 dblp:conf/cvpr/WangCYLH16 fatcat:uhwbhfvabrdxnfj7frmwxqxghu

Towards Context-Agnostic Learning Using Synthetic Data [article]

Charles Jin, Martin Rinard
2021 arXiv   pre-print
On several standard benchmarks for real-world image classification, we achieve robust performance in the context-agnostic setting, with good generalization to real world domains, whereas training directly  ...  on real world data without our techniques yields classifiers that are brittle to perturbations of the background.  ...  Domain randomization for transferring deep neural networks from simulation to the real world. In 2017 IEEE/RSJ international conference on intelligent robots and systems (IROS), pages 23–30.  ... 
arXiv:2005.14707v3 fatcat:3lnb7rzqcjeh7jldywsvw3u2zi

Synthetic Document Generator for Annotation-free Layout Recognition [article]

Natraj Raman, Sameena Shah, Manuela Veloso
2021 arXiv   pre-print
We empirically illustrate that a deep layout detection model trained purely on the synthetic documents can match the performance of a model that uses real documents.  ...  The proposed generative process treats every physical component of a document as a random variable and models their intrinsic dependencies using a Bayesian Network graph.  ...  intended as investment research or investment advice, or a recommendation, offer or solicitation for the purchase or sale of any security, financial instrument, financial product or service, or to be used  ... 
arXiv:2111.06016v1 fatcat:zgjkmt4z7vcr7ku4otagz7ljxa

Domain Generalization by Solving Jigsaw Puzzles

Fabio M. Carlucci, Antonio D'Innocente, Silvia Bucci, Barbara Caputo, Tatiana Tommasi
2019 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
Multiple experiments on the PACS, VLCS, Office-Home and digits datasets confirm our intuition and show that this simple method outperforms previous domain generalization and adaptation solutions.  ...  In this paper we propose to apply a similar approach to the task of object recognition across domains: our model learns the semantic labels in a supervised fashion, and broadens its understanding of the  ...  Acknowledgments This work was supported by the ERC grant 637076 RoboExNovo and a NVIDIA Academic Hardware Grant.  ... 
doi:10.1109/cvpr.2019.00233 dblp:conf/cvpr/CarlucciDBCT19 fatcat:oxii66di2bgzhn4m6qqccwsezq

Empowering Things with Intelligence: A Survey of the Progress, Challenges, and Opportunities in Artificial Intelligence of Things [article]

Jing Zhang, Dacheng Tao
2020 arXiv   pre-print
Artificial intelligence (AI), especially deep learning, is now a proven success in various areas including computer vision, speech recognition, and natural language processing.  ...  Next, we summarize some promising applications of AIoT that are likely to profoundly reshape our world. Finally, we highlight the challenges facing AIoT and some potential research opportunities.  ...  Integrated Circuit ASR Automatic Speech Recognition CNN Convolutional Neural Networks CTC Connectionist Temporal Classification DA Domain Adaptation DNN Deep Neural Network DRL Deep Reinforcement  ... 
arXiv:2011.08612v1 fatcat:dflut2wdrjb4xojll34c7daol4

Deep Domain-Adversarial Image Generation for Domain Generalisation [article]

Kaiyang Zhou, Yongxin Yang, Timothy Hospedales, Tao Xiang
2020 arXiv   pre-print
Specifically, DDAIG consists of three components, namely a label classifier, a domain classifier and a domain transformation network (DoTNet).  ...  In this paper, we propose a novel DG approach based on Deep Domain-Adversarial Image Generation (DDAIG).  ...  Method Artistic Clipart Product Real World Avg.  ... 
arXiv:2003.06054v1 fatcat:36sfvvrg6beqvd5na6lt4uivzy

Self-Supervised Noisy Label Learning for Source-Free Unsupervised Domain Adaptation [article]

Weijie Chen and Luojun Lin and Shicai Yang and Di Xie and Shiliang Pu and Yueting Zhuang and Wenqi Ren
2021 arXiv   pre-print
It is a strong prerequisite to access source data freely in many existing unsupervised domain adaptation approaches.  ...  Usually, the given source domain pre-trained model is expected to optimize with only unlabeled target data, which is termed as source-free unsupervised domain adaptation.  ...  This is the main factor of model performance degradation in real-world scenarios. Recently, there are more and more researchers delving into unsuper- Figure 1 .  ... 
arXiv:2102.11614v1 fatcat:noeiqof42bhzfaaxitkmaotvaq

Domain Adaptation with Incomplete Target Domains [article]

Zhenpeng Li, Jianan Jiang, Yuhong Guo, Tiantian Tang, Chengxiang Zhuo, Jieping Ye
2020 arXiv   pre-print
We conduct experiments on both cross-domain benchmark tasks and a real world adaptation task with imperfect target domains.  ...  However, the standard domain adaptation has assumed perfectly observed data in both domains, while in real world applications the existence of missing data can be prevalent.  ...  Experiment We conducted experiments on both benchmark digit recognition datasets for domain adaptation with simulated incomplete target domains and a real world domain adaptation problem with natural incomplete  ... 
arXiv:2012.01606v1 fatcat:po6r6icfrneavm2dpt5isnxyz4

Chargrid-OCR: End-to-end Trainable Optical Character Recognition for Printed Documents using Instance Segmentation [article]

Christian Reisswig, Anoop R Katti, Marco Spinaci, Johannes Höhne
2020 arXiv   pre-print
, therefore, significantly faster and (iii) is easy to train and adapt to a new domain.  ...  For training the model, we build two large-scale datasets without resorting to any manual annotation - synthetic documents with clean labels and real documents with noisy labels.  ...  To more closely mimic a real-world dataset, we perform data augmentation.  ... 
arXiv:1909.04469v4 fatcat:nux52deskvgnrmzzhq3qbrsmnq
« Previous Showing results 1 — 15 out of 6,383 results