Filters








94,599 Hits in 10.7 sec

MARGIN AND DOMAIN INTEGRATED CLASSIFICATION FOR IMAGES

YEN-LUN CHEN, YUAN F. ZHENG, YI LIU
2011 International Journal of Information Acquisition  
Multi-category classification is an on going research topic in image acquisition and processing for numerous applications.  ...  In this paper, a novel approach called margin and domain integrated classifier (MDIC) is addressed.  ...  To integrate margin and domain in a sin-gle model for multiple classification, we extend the capabilities of MEMEM to a much powerful tool, which is called margin and domain integrated classification (  ... 
doi:10.1142/s0219878911002343 fatcat:qzg4pojt25fc7k4k7b3ii2eh2a

Dynamic Integration with Random Forests [chapter]

Alexey Tsymbal, Mykola Pechenizkiy, Pádraig Cunningham
2006 Lecture Notes in Computer Science  
We conduct experiments on a selection of classification datasets, analysing the resulting accuracy, the margin and the bias and variance components of error.  ...  The experiments demonstrate that dynamic integration increases accuracy on some datasets. Even if the accuracy remains the same, dynamic integration always increases the margin.  ...  Analysis of Classification Margin Besides the accuracy for the different integration techniques considered we also measured margin for static voting, DV and DVS.  ... 
doi:10.1007/11871842_82 fatcat:xwnnz5lwvvdkdiouvccet2bnma

Class Distribution Alignment for Adversarial Domain Adaptation [article]

Wanqi Yang, Tong Ling, Chengmei Yang, Lei Wang, Yinghuan Shi, Luping Zhou, Ming Yang
2020 arXiv   pre-print
Furthermore, our approach enforces the classification consistence of target domain images before and after adaptation to aid the classifier training in both domains.  ...  Most existing unsupervised domain adaptation methods mainly focused on aligning the marginal distributions of samples between the source and target domains.  ...  Specifically, we compare the real and generated images for CycleGAN, SBADA and CADIT, and compare the generated images from both domains for ADGAN.  ... 
arXiv:2004.09403v1 fatcat:73m2iryit5dsfn2ccrz5dtwccm

Filter-Invariant Image Classification on Social Media Photos

Yu-Hsiu Chen, Ting-Hsuan Chao, Sheng-Yi Bai, Yen-Liang Lin, Wen-Chin Chen, Winston H. Hsu
2015 Proceedings of the 23rd ACM international conference on Multimedia - MM '15  
Convolutional Neural Network (CNN) has been shown as the state-of-the-art approach for image classification.  ...  To understand the image content, image classification becomes a very essential technique for plenty of applications (e.g., object detection, image caption generation).  ...  We treat original images as source domain and both original and filtered images as target domain to do supervised domain adaptation.  ... 
doi:10.1145/2733373.2806348 dblp:conf/mm/ChenCBLCH15 fatcat:gqusmuw5crgohjg6jykvoepznu

Deep Transfer Learning for Image Emotion Analysis: Reducing Marginal and Joint Distribution Discrepancies Together

Yuwei He, Guiguang Ding
2019 Neural Processing Letters  
The method can leverage rich emotion knowledge from a source domain to the target domain. Our method reduces both marginal and joint domain distribution discrepancies at fully-connected layers.  ...  Therefore, a CNN is hard to perform well in an image domain with scant labeled information. In this paper, we propose a deep transfer learning method for image emotion analysis.  ...  both MMD and JMMD into the FC layers of the CNN, where MMD is used for measuring marginal discrepancy and JMMD is used for measuring joint discrepancy for two domain.  ... 
doi:10.1007/s11063-019-10035-7 fatcat:kg6tndto5fdqxhslnxn6hd5qcu

New cascade model for hierarchical joint classification of multitemporal, multiresolution and multisensor remote sensing data

Ihsen Hedhli, Gabriele Moser, Josiane Zerubia, Sebastiano B. Serpico
2014 2014 IEEE International Conference on Image Processing (ICIP)  
In1this paper, we propose a novel method for the joint classification of multidate, multiresolution and multisensor remote sensing imagery, which represents a vital and fairly unexplored classification  ...  correlations associated with distinct images in the input time series.  ...  to integrate an exact estimator of the maximizer of posterior marginal (MPM) [8] [9] .  ... 
doi:10.1109/icip.2014.7026062 dblp:conf/icip/HedhliMZS14 fatcat:aj7qask3ujazzpmj647plhtm4e

Domain Invariant Representation Learning with Domain Density Transformations [article]

A. Tuan Nguyen, Toan Tran, Yarin Gal, Atılım Güneş Baydin
2022 arXiv   pre-print
To tackle this problem, a predominant approach is to find and learn some domain-invariant information in order to use it for the prediction task.  ...  Naively training a model on the aggregate set of data (pooled from all source domains) has been shown to perform suboptimally, since the information learned by that model might be domain-specific and generalize  ...  d . • The adversarial loss L adv that is the classification loss of a discriminator D that tries to distinguish between real images and the synthetic images generated by G.  ... 
arXiv:2102.05082v3 fatcat:dainw2whyrfj7kgt5f7hpcc67y

A Survey on Incorporating Domain Knowledge into Deep Learning for Medical Image Analysis [article]

Xiaozheng Xie, Jianwei Niu, Xuefeng Liu, Zhengsu Chen, Shaojie Tang, Shui Yu
2020 arXiv   pre-print
For each task, we systematically categorize different kinds of medical domain knowledge that have been utilized and their corresponding integrating methods.  ...  In this survey, we summarize the current progress on integrating medical domain knowledge into deep learning models for various tasks, such as disease diagnosis, lesion, organ and abnormality detection  ...  How to design an effective and useful integrating method is essential for medical image analysis.  ... 
arXiv:2004.12150v3 fatcat:2cqumcjkizgivmo67reznxacie

A survey of transfer learning

Karl Weiss, Taghi M. Khoshgoftaar, DingDing Wang
2016 Journal of Big Data  
Therefore, there is a need to create a high-performance learner for a target domain trained from a related source domain. This is the motivation for transfer learning.  ...  Lastly, there is information listed on software downloads for various transfer learning solutions and a discussion of possible future research work.  ...  The experiments are performed for the applications of image classification and text classification. The source contains labeled image data and the target contains limited labeled image data.  ... 
doi:10.1186/s40537-016-0043-6 fatcat:auxq6aafwfhgtjaofb5lf45v4u

Few-Shot Learning by Integrating Spatial and Frequency Representation [article]

Xiangyu Chen, Guanghui Wang
2021 arXiv   pre-print
We employ Discrete Cosine Transformation (DCT) to generate the frequency representation, then, integrate the features from both the spatial domain and frequency domain for classification.  ...  The classification accuracy is boosted significantly by integrating features from both the spatial and frequency domains in different few-shot learning tasks.  ...  ACKNOWLEDGEMENT The work was supported in part by The National Aeronautics and Space Administration (NASA) under grant no. 80NSSC20M0160.  ... 
arXiv:2105.05348v2 fatcat:rjro23latzftfdko5skw7h5eae

Discriminative Transfer Subspace Learning via Low-Rank and Sparse Representation

Yong Xu, Xiaozhao Fang, Jian Wu, Xuelong Li, David Zhang
2016 IEEE Transactions on Image Processing  
Index Terms-Source domain, target domain, low-rank and sparse constraints, knowledge transfer, subspace learning.  ...  In this way, the discrepancy of the source and target domains is reduced.  ...  Two key factors may contribute to the good classification performance for GFK: 1) the domain shift between these two domains can be well modeled by using the kernel that integrates all the subspaces along  ... 
doi:10.1109/tip.2015.2510498 pmid:26701675 fatcat:fl5arjzwwjdyjb7ty3ovtb5hmm

Recent Advances in Selection Techniques for Image Processing

Sathiyaraj Chinnasamy, M Ramachandran, Vidhya Prasanth
2022 Electrical and Automation Engineering  
So, it is not surprising that there are so many Image processing algorithms for margin extraction, upgrade, rearrangement; data compression, etc. are unambiguous.  ...  These usually include an introduction to the package and an insight, for image processing ideas Provides introductions.  ...  Used for an SVM classification and CNN for feature extraction in Huynh and Geiger.  ... 
doi:10.46632/eae/1/2/5 fatcat:h6yvghysizg7zdhtn323lz3klm

Transfer Feature Learning with Joint Distribution Adaptation

Mingsheng Long, Jianmin Wang, Guiguang Ding, Jiaguang Sun, Philip S. Yu
2013 2013 IEEE International Conference on Computer Vision  
Extensive experiments verify that JDA can significantly outperform several state-of-the-art methods on four types of cross-domain image classification problems.  ...  However, most prior methods have not simultaneously reduced the difference in both the marginal distribution and conditional distribution between domains.  ...  This verifies that JDA can construct more effective and robust representation for cross-domain image classification tasks.  ... 
doi:10.1109/iccv.2013.274 dblp:conf/iccv/LongWDSY13 fatcat:6k73a3sx2ffgjhykr7urppjnum

Explainable Supervised Domain Adaptation [article]

Vidhya Kamakshi, Narayanan C Krishnan
2022 arXiv   pre-print
We integrate a case-based reasoning mechanism into the XSDA-Net to explain the prediction of a test instance in terms of similar-looking regions in the source and target train images.  ...  Leveraging knowledge from an auxiliary source domain for learning in labeled data-scarce target domain is fundamental to domain adaptation.  ...  The source and target domains differ in the underlying marginal and conditional distributions.  ... 
arXiv:2205.09943v2 fatcat:ics7ezo4hnh5rkmlqc2kpvfwyq

Data Augmentation in Emotion Classification Using Generative Adversarial Networks [article]

Xinyue Zhu, Yifan Liu, Zengchang Qin, Jiahong Li
2017 arXiv   pre-print
It can complement and complete the data manifold and find better margins between neighboring classes.  ...  It is a difficult task to classify images with multiple class labels using only a small number of labeled examples, especially when the label (class) distribution is imbalanced.  ...  Experiments on three benchmark datasets show that our GAN-based data augmentation techniques can lead to improvement in distribution integrity and margin clarity between classes, and can obtain 5%∼10%  ... 
arXiv:1711.00648v5 fatcat:jcofsmztk5hlrnr3ypbocveya4
« Previous Showing results 1 — 15 out of 94,599 results