756 Hits in 9.7 sec

Cooperative Training and Latent Space Data Augmentation for Robust Medical Image Segmentation [article]

Chen Chen, Kerstin Hammernik, Cheng Ouyang, Chen Qin, Wenjia Bai, Daniel Rueckert
2021 arXiv   pre-print
Using only 10 subjects from a single site for training, we demonstrated improved cross-site segmentation performance and increased robustness against various unforeseen imaging artifacts compared to strong  ...  Both contributions improve model generalization and robustness with limited data. The cooperative training framework consists of a fast-thinking network (FTN) and a slow-thinking network (STN).  ...  Learning robust networks from single-domain data and limited data is of great practical value for medical imaging research.  ... 
arXiv:2107.01079v1 fatcat:dagl3yn32jh2dpvks525j4hkca

Robustness Testing of Data and Knowledge Driven Anomaly Detection in Cyber-Physical Systems [article]

Xugui Zhou, Maxfield Kouzel, Homa Alemzadeh
2022 arXiv   pre-print
Although robustness testing of deep learning models has been extensively explored in applications such as image classification and speech recognition, less attention has been paid to ML-driven safety monitoring  ...  generated using a Gaussian-based noise model and the Fast Gradient Sign Method (FGSM).  ...  White-box Attacks: Fast Gradient Sign Method (FGSM) [12] is a simple but effective method widely used in generating adversarial images using the gradients of a neural network, and is reported to be also  ... 
arXiv:2204.09183v2 fatcat:psgjp27oevckznsoxj6smxgmse

Improved Adversarial Robustness via Logit Regularization Methods [article]

Cecilia Summers, Michael J. Dinneen
2019 arXiv   pre-print
adversarial robustness at little to no marginal cost.  ...  In this paper, we advocate for and experimentally investigate the use of a family of logit regularization techniques as an adversarial defense, which can be used in conjunction with other methods for creating  ...  Investigating this phenomenon further, we examine two alternatives for logit regularization, finding that both result in improved robustness to adversarial examples, sometimes surprisingly so -for example  ... 
arXiv:1906.03749v1 fatcat:vg4pd5divndwbbgrywr7zy5ili

Robustness study of noisy annotation in deep learning based medical image segmentation [article]

Shaode Yu, Erlei Zhang, Junjie Wu, Hang Yu, Zi Yang, Lin Ma, Mingli Chen, Xuejun Gu, Weiguo Lu
2020 arXiv   pre-print
Partly due to the use of exhaustive-annotated data, deep networks have achieved impressive performance on medical image segmentation.  ...  This study suggests that deep learning-based medical image segmentation is robust to noisy annotations to some extent. It also highlights the importance of labeling quality in deep learning  ...  to the development of deep networks robustness and may be generalizable to noisy annotation.  ... 
arXiv:2003.06240v1 fatcat:rvpnezcp5zd3fnogvr7zn647ju

Towards Adversarially Robust Deep Image Denoising [article]

Hanshu Yan, Jingfeng Zhang, Jiashi Feng, Masashi Sugiyama, Vincent Y. F. Tan
2022 arXiv   pre-print
This work systematically investigates the adversarial robustness of deep image denoisers (DIDs), i.e, how well DIDs can recover the ground truth from noisy observations degraded by adversarial perturbations  ...  Firstly, to evaluate DIDs' robustness, we propose a novel adversarial attack, namely Observation-based Zero-mean Attack (ObsAtk), to craft adversarial zero-mean perturbations on given noisy images.  ...  For the PolyU and CC, we use the clean images in BSD500 for training an adversarially robust Table B .  ... 
arXiv:2201.04397v2 fatcat:wtuccgi2ivgszg6rbponld374i

On Fast Adversarial Robustness Adaptation in Model-Agnostic Meta-Learning [article]

Ren Wang, Kaidi Xu, Sijia Liu, Pin-Yu Chen, Tsui-Wei Weng, Chuang Gan, Meng Wang
2021 arXiv   pre-print
In addition to generalization, robustness is also desired for a meta-model to defend adversarial examples (attacks).  ...  Toward promoting adversarial robustness in MAML, we first study WHEN a robustness-promoting regularization should be incorporated, given the fact that MAML adopts a bi-level (fine-tuning vs. meta-update  ...  ACKNOWLEDGEMENT This work was supported by the Rensselaer-IBM AI Research Collaboration (http://airc., part of the IBM AI Horizons Network (  ... 
arXiv:2102.10454v1 fatcat:itferwngljd73fuxkij7xuczbe

Learning Image Labels On-the-fly for Training Robust Classification Models [article]

Xiaosong Wang, Ziyue Xu, Dong Yang, Leo Tam, Holger Roth, Daguang Xu
2020 arXiv   pre-print
Multi-observer studies have been conducted to study these annotation variances (by labeling the same data for multiple times) and its effects on critical applications like medical image analysis.  ...  On the other hand, automated annotation methods based on NLP algorithms have recently shown promise as a reasonable alternative, relying on the existing diagnostic reports of those images that are widely  ...  The pairs are defined as Airplane vs Bird, cat vs dog, dear vs cat, horse vs deer, ship vs airplane, and truck vs automobile.  ... 
arXiv:2009.10325v2 fatcat:3fddd7zdurbc5gmkqokxuhdhpi

Robust deep learning classification of adamantinomatous craniopharyngioma from limited preoperative radiographic images

Eric W. Prince, Ros Whelan, David M. Mirsky, Nicholas Stence, Susan Staulcup, Paul Klimo, Richard C. E. Anderson, Toba N. Niazi, Gerald Grant, Mark Souweidane, James M. Johnston, Eric M. Jackson (+15 others)
2020 Scientific Reports  
Herein, we present a novel genetic algorithm (GA) that identifies optimal architecture parameters using feature embeddings from state-of-the-art image classification networks to identify the pediatric  ...  Such an approach will depend upon highly accurate models built using the limited datasets that are available.  ...  The second method, known as TANDA (Transformation Adversarial Networks for Data Augmentation), is a ML-based approach that uses Generative Adversarial Networks (GANs) and Recurrent Neural Network (RNNs  ... 
doi:10.1038/s41598-020-73278-8 pmid:33037266 fatcat:6ziupgjdynejnkw473ll6y6usy

Adaptive Boosting for Domain Adaptation: Towards Robust Predictions in Scene Segmentation [article]

Zhedong Zheng, Yi Yang
2021 arXiv   pre-print
We have achieved competitive results on three widely-used scene segmentation domain adaptation benchmarks.  ...  Extensive experiments show that (1) Without the need to worry about the stopping time, AdaBoost Student provides one robust solution by efficient complementary model learning during training. (2) AdaBoost  ...  In the future, we will continue to investigate the usage of Adaboost Student and apply it to other fields, such as medical images, to obtain the model with good generalizability. Fig. 1 . 1 Fig. 1.  ... 
arXiv:2103.15685v2 fatcat:ju5bi4w36jawzpeqpr2ksxuf4q

Adversarial Attack Vulnerability of Deep Learning Models for Oncologic Images [article]

Marina Z. Joel, Sachin Umrao, Enoch Chang, Rachel Choi, Daniel X Yang, Antonio Omuro, Roy Herbst, Harlan Krumholz, Sanjay Aneja
2021 medRxiv   pre-print
We investigated how PGD adversarial training could be employed to increase model robustness against FGSM, PGD, and BIM attacks.  ...  Our findings provide a useful basis for designing more robust and accurate medical DL models as well as techniques to defend models from adversarial attack.  ...  The proposed networks were implemented in Python 2.7 using TensorFlow v1.15.3 framework (26) . Adversarial images were created using the Adversarial Robustness Toolbox v1.4.1 (27) .  ... 
doi:10.1101/2021.01.17.21249704 fatcat:kcwjm772ufcg3h7zh6fpujyzxa

Robust contour propagation using deep learning and image registration for online adaptive proton therapy of prostate cancer

Mohamed S. Elmahdy, Thyrza Jagt, R. Th. Zinkstok, Yuchuan Qiao, Rahil Shahzad, Hessam Sokooti, Sahar Yousefi, Luca Incrocci, C.A.M. Marijnen, Mischa Hoogeman, Marius Staring
2019 Medical Physics (Lancaster)  
To develop and validate a robust and accurate registration pipeline for automatic contour propagation for online adaptive Intensity-Modulated Proton Therapy (IMPT) of prostate cancer using elastix software  ...  The bladder segmentation network achieved a DSC of 88% and 82% on the test datasets.  ...  ACKNOWLEDGMENTS This study was financially supported by Varian Medical Systems and ZonMw, the Netherlands Organization for Health Research and Development, grant number 104003012.  ... 
doi:10.1002/mp.13620 pmid:31111962 pmcid:PMC6852565 fatcat:jbm4msfaqzgpzk2deibvvvmkba

From Characters to Understanding Natural Language (C2NLU): Robust End-to-End Deep Learning for NLP (Dagstuhl Seminar 17042)

Phil Blunsom, Kyunghyun Cho, Chris Dyer, Hinrich Schütze, Marc Herbstritt
2017 Dagstuhl Reports  
This report documents the program and the outcomes of Dagstuhl Seminar 17042 "From Characters to Understanding Natural Language (C2NLU): Robust End-to-End Deep Learning for NLP".  ...  natural language processing, computational linguistics, deep learning and general machine learning. 31 participants from 22 academic and industrial institutions discussed advantages and challenges of using  ...  Investigation of BPE (offline segmentation) vs. character-based (online Example for data is very large (infinite).  ... 
doi:10.4230/dagrep.7.1.129 dblp:journals/dagstuhl-reports/BlunsomCDS17 fatcat:lyp7srzsg5cgngiklccjox4abm

A 3D Fully Convolutional Neural Network With Top-Down Attention-Guided Refinement for Accurate and Robust Automatic Segmentation of Amygdala and Its Subnuclei

Yilin Liu, Brendon M. Nacewicz, Gengyan Zhao, Nagesh Adluru, Gregory R. Kirk, Peter A. Ferrazzano, Martin A. Styner, Andrew L. Alexander
2020 Frontiers in Neuroscience  
We also demonstrated the feasibility of using a cycle-consistent generative adversarial network (CycleGAN) to harmonize multi-site MRI data, and show that our method generalizes well to challenging traumatic  ...  This appears to be a promising strategy for image segmentation for multiple site studies and increased morphological variability from significant brain pathology.  ...  As there was considerable site-to-site variability, we investigated the utility of a cycle-consistent generative adversarial network approach (CycleGAN) to harmonize the image contrast with the training  ... 
doi:10.3389/fnins.2020.00260 pmid:32508558 pmcid:PMC7253589 fatcat:afxc5co2xfdh3no5alnkhm6h34

Correct-N-Contrast: A Contrastive Approach for Improving Robustness to Spurious Correlations [article]

Michael Zhang, Nimit S. Sohoni, Hongyang R. Zhang, Chelsea Finn, Christopher Ré
2022 arXiv   pre-print
Spurious correlations pose a major challenge for robust machine learning.  ...  a robust model with contrastive learning to learn similar representations for same-class samples.  ...  For example, consider classifying cows versus camels in natural images. 90% of cows may appear on grass and 90% of camels on sand.  ... 
arXiv:2203.01517v1 fatcat:iqs5btv2nrdwzhylocwdoh4oqq

Validation and Generalizability of Self-Supervised Image Reconstruction Methods for Undersampled MRI [article]

Thomas Yu, Tom Hilbert, Gian Franco Piredda, Arun Joseph, Gabriele Bonanno, Salim Zenkhri, Patrick Omoumi, Meritxell Bach Cuadra, Erick Jorge Canales-Rodríguez, Tobias Kober, Jean-Philippe Thiran
2022 arXiv   pre-print
Two self-supervised algorithms based on self-supervised denoising and the deep image prior were investigated.  ...  We further showed that no-reference image metrics correspond well with human rating of image quality for studying generalizability.  ...  They have been shown to potentially be useful for MR/medical image evaluation without ground truth (27; 28); we use the following three metrics: a metric used originally for assessing the quality of JPEG-compressed  ... 
arXiv:2201.12535v2 fatcat:tewdsu3scba5rdz5kfiixtcmyu
« Previous Showing results 1 — 15 out of 756 results