491,794 Hits in 5.0 sec

On Feature Normalization and Data Augmentation [article]

Boyi Li and Felix Wu and Ser-Nam Lim and Serge Belongie and Kilian Q. Weinberger
2021 arXiv   pre-print
As our approach is fast, operates entirely in feature space, and mixes different signals than prior methods, one can effectively combine it with existing augmentation approaches.  ...  in addition to the normalized features.  ...  of Naval Research DOD (N00014-17-1-2175), Bill and Melinda Gates Foundation.  ... 
arXiv:2002.11102v3 fatcat:vxcuukrj6zgyjp4k4xk3bbboru

Feature Extension of Gut Microbiome Data for Deep Neural Network Based Colorectal Cancer Classification

Mwenge Mulenga, Sameem Abdul Kareem, Aznul Qalid Md Sabri, Manjeevan Seera, Suresh Govind, Chandramathi Samudi, Saharuddin Mohamad
2021 IEEE Access  
In this paper, we propose a feature augmentation approach that aggregates data normalization methods to extend existing features of a dataset.  ...  The proposed method combines feature extension with data augmentation to improve CRC classification performance of a DNN model.  ...  on both nonaugmented and augmented data.  ... 
doi:10.1109/access.2021.3050838 fatcat:ikqjdlpnffanxbxjnbijligx5q

Exploit Direction Information for Remote Ship Detection

Zhenbiao Tan, Zekun Zhang, Tingzhuang Xing, Xiao Huang, Junbin Gong, Jie Ma
2021 Remote Sensing  
On the L1 task of the HRSC2016 data set, the direction augmentation method and direction normalization method boost the RoI Transformer baseline from 86.2% to 90.4% and 90.6%, respectively, achieving the  ...  The direction augmentation method directly augments the features of ship RRoIs and brings great diversities to the training data set.  ...  On the L1 task of the HRSC2016 data set, the proposed direction augmentation and the direction normalization methods both achieved the state-of-the-art performance.  ... 
doi:10.3390/rs13112155 fatcat:elz7n5yajzhgjiosvcdpbbmh2y

Fundamental Frequency Feature Normalization and Data Augmentation for Child Speech Recognition [article]

Gary Yeung, Ruchao Fan, Abeer Alwan
2021 arXiv   pre-print
This study proposes a novel technique for child ASR using both feature normalization and data augmentation methods based on the relationship between formants and fundamental frequency (f_o).  ...  Both the f_o feature normalization and data augmentation techniques are implemented as a frequency shift in the Mel domain. These techniques are evaluated on a child read speech ASR task.  ...  Another strategy is to augment the training data by creating additional speech-like features for training data.  ... 
arXiv:2102.09106v1 fatcat:huz3cj67tnhxjnx23dwj3lhrgm

Significance of Data Augmentation for Improving Cleft Lip and Palate Speech Recognition [article]

Protima Nomo Sudro, Rohan Kumar Das, Rohit Sinha, S. R. Mahadeva Prasanna
2021 arXiv   pre-print
Our study finds that the data augmentation methods significantly improve the CLP speech recognition performance, which is more evident when we used feature modification using CycleGAN, VTLP and reverberation  ...  In line with the challenge, in this work, we investigate a few data augmentation techniques to simulate training data for improving the children speech recognition considering the case of cleft lip and  ...  Pushpavathi and Dr. Ajish Abraham, AIISH Mysore, for providing insights about CLP speech disorder.  ... 
arXiv:2110.00797v1 fatcat:nsv6icgf6bfgjebvio3lcfmcsa

AugMax: Adversarial Composition of Random Augmentations for Robust Training [article]

Haotao Wang, Chaowei Xiao, Jean Kossaifi, Zhiding Yu, Anima Anandkumar, Zhangyang Wang
2022 arXiv   pre-print
To solve this problem, we further design a disentangled normalization module, termed DuBIN (Dual-Batch-and-Instance Normalization), that disentangles the instance-wise feature heterogeneity arising from  ...  Diversity and hardness are two complementary dimensions of data augmentation to achieve robustness.  ...  Specifically, we visualize the impact of various data augmentation methods on the feature representations obtained with a ResNeXt29 that is normally trained on CIFAR-10.  ... 
arXiv:2110.13771v3 fatcat:e7ulbwviprhyzpoyt72m2tlzem

Understanding Unconventional Preprocessors in Deep Convolutional Neural Networks for Face Identification [article]

Chollette C. Olisah, Lyndon Smith
2019 arXiv   pre-print
Commonly used preprocessors are the data augmentation and normalization and are termed conventional preprocessors.  ...  However, for best performance, the right setup of preprocessed data with augmentation and/or normalization is required.  ...  Additionally, it is obvious that different preprocessors react differently with data augmentation and normalization. The RGB, HE, CFSP, and LN perform better when data are normalized and augmented.  ... 
arXiv:1904.00815v2 fatcat:bemcyxjyube7fmbtnqhcgdxpvy

Towards Fair Cross-Domain Adaptation via Generative Learning [article]

Tongxin Wang, Zhengming Ding, Wei Shao, Haixu Tang, Kun Huang
2020 arXiv   pre-print
Specifically, generative feature augmentation is explored to synthesize effective training data for few-shot source classes, while effective cross-domain alignment aims to adapt knowledge from source to  ...  Experimental results on two large cross-domain visual datasets demonstrate the effectiveness of our proposed method on improving both few-shot and overall classification accuracy comparing with the state-of-the-art  ...  Data Augmentation Data augmentation is a straightforward approach to improve the performance on few-shot classification.  ... 
arXiv:2003.02366v2 fatcat:73q2wegggjhixo462wya2czore

Self-Supervised Learning for Anomaly Detection with Dynamic Local Augmentation

Seungdong Yoa, Seungjun Lee, Chiyoon KIM, Hyunwoo J. Kim
2021 IEEE Access  
Our experiment demonstrates the effectiveness of our method, and we show that our framework achieves competitive performance compared to state-of-the-art methods on MVTec Anomaly Detection dataset.  ...  Specifically, in addition to learning the global representation of an image, our framework contrasts a normal sample to a locally augmented sample.  ...  only the normal data.  ... 
doi:10.1109/access.2021.3124525 fatcat:hl46mb5llzeo3jsebuujmm76ky

Histopathological Classification of Breast Cancer Images Using a Multi-Scale Input and Multi-Feature Network

Taimoor Shakeel Sheikh, Yonghee Lee, Migyung Cho
2020 Cancers  
The MSI-MFNet predicts the probability of a disease on the patch and image levels. We evaluated the performance of our proposed model on two public benchmark datasets.  ...  We propose a multi-scale input and multi-feature network (MSI-MFNet) model, which can learn the overall structures and texture features of different scale tissues by fusing multi-resolution hierarchical  ...  WDA (without data augmentation) and DA (data augmentation) Table 7 .  ... 
doi:10.3390/cancers12082031 pmid:32722111 pmcid:PMC7465368 fatcat:j6r67tmzh5brnlvpssejjjxlfq

NADS-RA: Network Anomaly Detection Scheme Based on feature Representation and data Augmentation

Xu Liu, Xiaoqiang Di, Qiang Ding, Weiyou Liu, Hui Qi, Jinqing Li, Huamin Yang
2020 IEEE Access  
Therefore, this paper addresses this issue by proposing a Network Anomaly Detection Scheme based on feature Representation and data Augmentation (NADS-RA).  ...  We conduct experiments on five public benchmark datasets, including NSL-KDD and UNSW-NB15, and so on, and compare against 12 detection methods and 17 data generation methods.  ...  Data preprocessing includes feature encoding, feature reduction and normalization. Feature representation is performed on training, validation and test datasets.  ... 
doi:10.1109/access.2020.3040510 fatcat:muqfbnj6kragdg5ui36afajxuy

A Study On Data Augmentation In Voice Anti-Spoofing [article]

Ariel Cohen, Inbal Rimon, Eran Aflalo, Haim Permuter
2021 arXiv   pre-print
In addition, a new type of online data augmentation, SpecAverage, is introduced in which the audio features are masked with their average value in order to improve generalization.  ...  Our results are based on the ASVspoof 2021 challenge, in the Logical Access (LA) and Deep Fake (DF) categories.  ...  We chose models which had good performance on the ASVspoof 2019 data, and focused our efforts on data augmentation and feature design in order to tackle the challenges of ASVspoof 2021.  ... 
arXiv:2110.10491v1 fatcat:sr5vbobyfzgrhk3m4sspp7cixy

Understanding unconventional preprocessors in deep convolutional neural networks for face identification

Chollette C. Olisah, Lyndon Smith
2019 SN Applied Sciences  
Commonly used preprocessors are the data augmentation and normalization and are termed conventional preprocessors.  ...  However, for best performance, the right setup of preprocessed data with augmentation and/or normalization is required.  ...  The raw RGB, HE, CFSP, LN and LSSF perform better when data is normalized and augmented. The YCBCR is best with only data augmentation and rgbGELog is best with only normalized data.  ... 
doi:10.1007/s42452-019-1538-5 fatcat:cbowi6fwjzbi5mgo4ab3p52ovi

Do CNNs Encode Data Augmentations? [article]

Eddie Yan, Yanping Huang
2021 arXiv   pre-print
A fundamental question is whether neural network features encode data augmentation transformations.  ...  Our approach uses features in pre-trained vision models with minimal additional processing to predict common properties transformed by augmentation (scale, aspect ratio, hue, saturation, contrast, and  ...  Which data augmentations correspond to low-level model features, and which correspond to high-level model features?  ... 
arXiv:2003.08773v3 fatcat:ultgzwmrondtvimnh5gucqvlma

An Image Clustering Auto-Encoder Based on Predefined Evenly-Distributed Class Centroids and MMD Distance [article]

Qiuyu Zhu, Zhengyong Wang
2019 arXiv   pre-print
Specifically, we perform one-to-one data augmentation such as rotation, shear, and shift before data is input to the encoder to learn the more effective features.  ...  The data and the enhanced data are simultaneously input into the auto-encoder to obtain latent features and augmented latent features whose similarity are constrained by an augmentation loss.  ...  Specifically, one-to-one data augmentation is performed before the data is input to the encoder, and the data and the augmented data are simultaneously input to the encoder to obtain latent features and  ... 
arXiv:1906.03905v1 fatcat:wbjaha3gtzdzxefgua2nvhb5hu
« Previous Showing results 1 — 15 out of 491,794 results