Filters








6,237 Hits in 1.7 sec

Unsupervised Feature Learning via Non-Parametric Instance-level Discrimination [article]

Zhirong Wu, Yuanjun Xiong, Stella Yu, Dahua Lin
2018 arXiv   pre-print
Neural net classifiers trained on data with annotated class labels can also capture apparent visual similarity among categories without being directed to do so. We study whether this observation can be extended beyond the conventional domain of supervised learning: Can we learn a good feature representation that captures apparent similarity among instances, instead of classes, by merely asking the feature to be discriminative of individual instances? We formulate this intuition as a
more » ... ic classification problem at the instance-level, and use noise-contrastive estimation to tackle the computational challenges imposed by the large number of instance classes. Our experimental results demonstrate that, under unsupervised learning settings, our method surpasses the state-of-the-art on ImageNet classification by a large margin. Our method is also remarkable for consistently improving test performance with more training data and better network architectures. By fine-tuning the learned feature, we further obtain competitive results for semi-supervised learning and object detection tasks. Our non-parametric model is highly compact: With 128 features per image, our method requires only 600MB storage for a million images, enabling fast nearest neighbour retrieval at the run time.
arXiv:1805.01978v1 fatcat:3ilhvltoinfo5ckxynd5fhjtdq

The Emergence of Objectness: Learning Zero-Shot Segmentation from Videos [article]

Runtao Liu, Zhirong Wu, Stella X. Yu, Stephen Lin
2021 arXiv   pre-print
Humans can easily segment moving objects without knowing what they are. That objectness could emerge from continuous visual observations motivates us to model grouping and movement concurrently from unlabeled videos. Our premise is that a video has different views of the same scene related by moving components, and the right region segmentation and region flow would allow mutual view synthesis which can be checked from the data itself without any external supervision. Our model starts with two
more » ... eparate pathways: an appearance pathway that outputs feature-based region segmentation for a single image, and a motion pathway that outputs motion features for a pair of images. It then binds them in a conjoint representation called segment flow that pools flow offsets over each region and provides a gross characterization of moving regions for the entire scene. By training the model to minimize view synthesis errors based on segment flow, our appearance and motion pathways learn region segmentation and flow estimation automatically without building them up from low-level edges or optical flows respectively. Our model demonstrates the surprising emergence of objectness in the appearance pathway, surpassing prior works on zero-shot object segmentation from an image, moving object segmentation from a video with unsupervised test-time adaptation, and semantic image segmentation by supervised fine-tuning. Our work is the first truly end-to-end zero-shot object segmentation from videos. It not only develops generic objectness for segmentation and tracking, but also outperforms prevalent image-based contrastive learning methods without augmentation engineering.
arXiv:2111.06394v1 fatcat:fujfxghw2vfdphl6dtyafsjo4i

Open Compound Domain Adaptation [article]

Ziwei Liu, Zhongqi Miao, Xingang Pan, Xiaohang Zhan, Dahua Lin, Stella X. Yu, Boqing Gong
2020 arXiv   pre-print
A typical domain adaptation approach is to adapt models trained on the annotated data in a source domain (e.g., sunny weather) for achieving high performance on the test data in a target domain (e.g., rainy weather). Whether the target contains a single homogeneous domain or multiple heterogeneous domains, existing works always assume that there exist clear distinctions between the domains, which is often not true in practice (e.g., changes in weather). We study an open compound domain
more » ... n (OCDA) problem, in which the target is a compound of multiple homogeneous domains without domain labels, reflecting realistic data collection from mixed and novel situations. We propose a new approach based on two technical insights into OCDA: 1) a curriculum domain adaptation strategy to bootstrap generalization across domains in a data-driven self-organizing fashion and 2) a memory module to increase the model's agility towards novel domains. Our experiments on digit classification, facial expression recognition, semantic segmentation, and reinforcement learning demonstrate the effectiveness of our approach.
arXiv:1909.03403v2 fatcat:ax627aerszhxnk2fwziugjwdjy

A STELLA model for the estimation of atrazine runoff, leaching, adsorption, and degradation from an agricultural land

Ying Ouyang, Jia En Zhang, Dian Lin, Guo Dong Liu
2009 Journal of Soils and Sediments  
Recommendations and perspectives This study suggests that the model, developed with STELLA, has great  ...  The model was constructed using the commercially available package Structural Thinking, Experiential Learning Laboratory with Animation (STELLA) and calibrated using field data prior to its applications  ...  This condition was superimposed in the STELLA model.  ... 
doi:10.1007/s11368-009-0107-8 fatcat:zt6caaspajafbb6vupwmeu7bwq

Epistemic dialog between health services and operations research

Maria Stella de Castro Lobo, Marcos Pereira Estellita Lins
2010 Pesquisa Operacional  
Pesquisa Operacional, v.30, n.2, p.371-386, Maio a Agosto de 2010 Lobo & Lins -Epistemic dialog between health services and operations researchPesquisa Operacional, v.30, n.2, p.371-386, Maio a Agosto  ... 
doi:10.1590/s0101-74382010000200007 fatcat:d5phttn3ebd4jeb3b2vi4qeeqy

Prolonged Exposure to Air Pollution Increases Periodontal Disease Risk: A Nationwide, Population-Based, Cohort Study

Han-Jie Lin, Stella Chin-Shaw Tsai, Frank Cheau-Feng Lin, Yi-Chao Hsu, Shih-Wei Chen, Ruey-Hwang Chou, Cheng-Li Lin, Chung Y. Hsu, Kuang-Hsi Chang
2021 Atmosphere  
(1) Background: No association between air pollution and periodontitis has yet been shown. Thus, we merged two nationwide databases to evaluate the risk of periodontitis in Taiwanese residents with long-term exposure to air pollution. (2) Methods: We conducted a nationwide retrospective cohort study using the Longitudinal Generation Tracking Database and the Taiwan Air Quality-Monitoring Database. The daily average air pollutant concentrations were categorized into quartiles (Q1, Q2, Q3, and
more » ... . We carried out Cox proportional hazards models to compute the hazard ratios of periodontitis, with 95% confidence intervals, in Q2–Q4 of the daily average air pollutant concentrations, compared with Q1. (3) Results: the adjusted HR (95 CI%) for periodontitis in Q2–Q4 increased with increased exposure to SO2, CO, NO, NO2, NOX, PM2.5, and PM10 from 1.72 (1.70, 1.76) to 4.86 (4.78–4.94); from 1.89 (1.85–1.93) to 2.64 (2.59–2.70); from 1.04 (1.02–1.06) to 1.52 (1.49–1.55); from 1.61 (1.58–1.64) to 2.51 (2.47–2.56); from 1.48 (1.45–1.51) to 2.11 (2.07–2.15); from 2.02 (1.98–2.06) to 22.9 (22.4–23.4, and from 2.71 (2.66–2.77) to 17.2 (16.8–17.6), respectively, compared to Q1. (4) Conclusions: Residents in Taiwan with long-term exposure to higher levels of air pollutants had a greater risk of periodontitis.
doi:10.3390/atmos12121668 fatcat:6engkbhhr5einofgeuv3t5h2ia

Computed tomography-guided localization with laser angle guide for thoracic procedures

Frank Cheau-Feng Lin, Stella Chin-Shaw Tsai, Hsien-Tang Tu, Yi-Lin Lai, Tzu-Chin Wu
2018 Journal of Thoracic Disease  
Computed tomography (CT)-guided lung procedures such as preoperative localizations, biopsies, and ablations are associated with morbidity even mortality. Often, the puncture angle is determined by the 'experienced hand' without a precise guide. We describe here the Laser Angle Guide Assembly ® to direct and steer the puncture angles precisely. It decreases procedure-related complications, saves time, reduces costs, avoids repeated punctures, and minimizes radiation exposures. TC. CT-guided
more » ... ization with laser angle guide for thoracic procedures.
doi:10.21037/jtd.2018.05.162 pmid:30069383 pmcid:PMC6051789 fatcat:ugi6lqeizzha5jeftncozj45si

3D Shape Reconstruction from Free-Hand Sketches [article]

Jiayun Wang, Jierui Lin, Qian Yu, Runtao Liu, Yubei Chen, Stella X. Yu
2022 arXiv   pre-print
Sketches are the most abstract 2D representations of real-world objects. Although a sketch usually has geometrical distortion and lacks visual cues, humans can effortlessly envision a 3D object from it. This suggests that sketches encode the information necessary for reconstructing 3D shapes. Despite great progress achieved in 3D reconstruction from distortion-free line drawings, such as CAD and edge maps, little effort has been made to reconstruct 3D shapes from free-hand sketches. We study
more » ... s task and aim to enhance the power of sketches in 3D-related applications such as interactive design and VR/AR games. Unlike previous works, which mostly study distortion-free line drawings, our 3D shape reconstruction is based on free-hand sketches. A major challenge for free-hand sketch 3D reconstruction comes from the insufficient training data and free-hand sketch diversity, e.g. individualized sketching styles. We thus propose data generation and standardization mechanisms. Instead of distortion-free line drawings, synthesized sketches are adopted as input training data. Additionally, we propose a sketch standardization module to handle different sketch distortions and styles. Extensive experiments demonstrate the effectiveness of our model and its strong generalizability to various free-hand sketches. Our code is publicly available at https://github.com/samaonline/3D-Shape-Reconstruction-from-Free-Hand-Sketches.
arXiv:2006.09694v2 fatcat:yt46gmnwmngztkngmo32lhlase

Selected Health Status Measures of Children from US Immigrant Families

Stella M. Yu, Sue C. Lin, Terry Adirim
2013 ISRN Pediatrics  
Using the 2007 National Survey of Children's Health (N = 91,532), we studied the relationship between the joint effects of immigrant family type (foreign-born children, US-born children/one foreign-born parent, US-born children/both foreign-born parents, and US-born children/US-born parents) and race/ethnicity on various health measures (parent-reported physical and dental health, obesity/overweight, breast-feeding, school absence, injury, and chronic condition). We used weighted logistic
more » ... sion to examine the independent effects of the 12-level joint variable on various health status measures while controlling for confounding factors. Overall, nearly one-third of families with both foreign-born parents were poor, and one-quarter of the parents in these households did not complete high school. Compared with non-Hispanic White US-born children, multivariable analyses indicate that all Hispanic children have higher odds of obesity, poor physical and dental health, with Hispanic foreign-born children 7 times as likely to report poor/fair physical health. Most children of immigrant parents were more likely to have been breast-fed and less likely to miss school more than 11 days. Child age and household poverty status were independently associated with most of the health status measures. Combined race/ethnicity and immigrant family type categories have heterogeneous associations with each health outcome measure examined. Culturally competent interventions and policies should be developed to serve these expanding communities.
doi:10.1155/2013/164757 pmid:23936667 pmcid:PMC3725827 fatcat:h4ocmcvfn5badgo27l6q57qv3a

Robot-assisted excision of cervical cystic hygroma through a retroauricular hairline approach: a case report

Frank Cheau-Feng Lin, Tsung-Lin Yang, Min-Che Tung, Stella Chin-Shaw Tsai
2016 Journal of Medical Case Reports  
Cystic hygroma is a rare benign abnormality of the lymphatic system generally occurring in young children less than 2 years old. The standard transcervical surgical treatment of cystic hygroma may often leave a permanent scar in the neck region. Case presentation: We report a case of cystic hygroma in a 19-month-old Asian baby girl successfully treated with robot-assisted excision through a hairline neck-lift approach. We present the use of the Yang's retractor as an instrumental advancement to
more » ... this surgical approach. Conclusions: Treatment options for cystic hygroma may be surgical or nonsurgical. We report a case of cystic hygroma in a 19-month-old child successfully treated with robot-assisted excision through a small concealed retroauricular hairline approach. This is the first report in the medical literature of treating cystic hygroma with a minimally invasive robot-assisted excision via a small, concealed, hairline incision.
doi:10.1186/s13256-016-0929-0 pmid:27276912 pmcid:PMC4899928 fatcat:5txazyqyxbel7hxezakckanlga

Exposure to Air Pollutants Increases the Risk of Chronic Rhinosinusitis in Taiwan Residents

Shih-Wei Chen, Han-Jie Lin, Stella Chin-Shaw Tsai, Cheng-Li Lin, Chung Y. Hsu, Tsai-Ling Hsieh, Chuan-Mu Chen, Kuang-Hsi Chang
2022 Toxics  
Air pollution triggers a tissue-specific inflammatory response. However, studies on the association between exposure to air pollutants and chronic rhinosinusitis (CRS) risk remain limited. Thus, we conducted this nationwide study to define the association between air pollution and CRS. We used the Longitudinal Health Insurance Database (LHID) and Taiwan Air Quality-Monitoring Database (TAQMD) to conduct a population-based cohort study. Study participants were recruited from the LHID, a data
more » ... et of the National Health Insurance Research Database that randomly sampled one million individuals. TAQMD has been an air pollutant database since 1998. In univariate and multivariate Cox proportional hazards regression models, adjusted hazard ratios (aHRs) and 95% CIs of CRS in five air pollutants were accounted. We adjusted for age, sex, urbanization level, insurance fee, comorbidities, and pollutant levels in the multivariate model. The total number of participants enrolled in this study was 160,504. The average age was 40.46 ± 14.62 years; males constituted 43.8% of the total participants. The percentages of alcoholism, tobacco dependence, and COPD were 1.5%, 2.8%, and 28.3%, respectively. After adjustment for age, sex, urbanization level, insurance fee, and comorbidities, the highest levels of air pollutants, including PM2.5 (aHR = 1.14, 95% CI = 1.06–1.22), NO2 (aHR = 1.07, 95% CI = 1.00–1.15), and PM10 (aHR = 1.13, 95% CI = 1.05–1.21) had a significantly greater CRS risk; we selected the lower concentration as the reference but did not correlate with CO. We found a significantly increased risk of CRS in residents with air pollutant exposure.
doi:10.3390/toxics10040173 pmid:35448434 pmcid:PMC9031629 fatcat:ba2bwic54vbphlg67dccheej5a

Effect of pulse depletion in a Brillouin optical time-domain analysis system

Luc Thévenaz, Stella Foaleng Mafang, Jie Lin
2013 Optics Express  
Energy transfer between the interacting waves in a distributed Brillouin sensor can result in a distorted measurement of the local Brillouin gain spectrum, leading to systematic errors. It is demonstrated that this depletion effect can be precisely modelled. This has been validated by experimental tests in an excellent quantitative agreement. Strict guidelines can be enunciated from the model to make the impact of depletion negligible, for any type and any length of fiber.
doi:10.1364/oe.21.014017 pmid:23787592 fatcat:246zfjxf3baxjhnzarbtgr6btu

Circumvention of multi-drug resistance of cancer cells by Chinese herbal medicines

Stella Chai, Kenneth KW To, Ge Lin
2010 Chinese Medicine  
Multi-drug resistance (MDR) of cancer cells severely limits therapeutic outcomes. A proposed mechanism for MDR involves the efflux of anti-cancer drugs from cancer cells, primarily mediated by ATP-binding cassette (ABC) membrane transporters including P-glycoprotein. This article reviews the recent progress of using active ingredients, extracts and formulae from Chinese medicine (CM) in circumventing ABC transporters-mediated MDR. Among the ABC transporters, Pgp is the most extensively studied
more » ... or its role in MDR reversal effects. While other MDR reversal mechanisms remain unclear, Pgp inhibition is a criterion for further mechanistic study. More mechanistic studies are needed to fully establish the pharmacological effects of potential MDR reversing agents. Review Multi-drug resistance (MDR) Multi-drug resistance (MDR) in cancer chemotherapy refers to the ability of cancer cells to survive from treatment of a wide range of drugs [1] . In addition to the MDR induced by drugs in early exposure, the MDR cancer cells may subsequently develop cross-resistance to several unexposed and structurally unrelated chemotherapeutic agents [2] . Mechanisms of MDR include decreased uptake of drugs, alterations in cellular pathways and increased active efflux of drugs [3] [4] [5] (Figure 1 ). Overexpression of ATP-binding cassette (ABC) transporters is one of the most common mechanisms. ABC transporters are large membrane-bound proteins consisting of two nucleotide-binding domains (NBDs) and two transmembrane domains (TMDs) which mediate the active transport of substrate drugs out of the cell (Figure 2) . Overexpression of the three major ABC transporters, i.e. P-glycoprotein (Pgp), multidrug-resistance-associated protein 1 (MRP1) and breast cancer resistance protein (BCRP/ABCG2) is frequently observed in cancer cell lines selected with chemotherapeutic drugs [6] and critical to clinical drug resistance [7] . P-glycoprotein (Pgp) P-glycoprotein (Pgp) [8] , which is also referred to as ABCB1 and MDR1, is the most studied ABC transporter. Pgp transports a wide range of chemotherapeutic
doi:10.1186/1749-8546-5-26 pmid:20653978 pmcid:PMC2920269 fatcat:yjoz5qxabbhatoq4kyljf2tcjq

Oesophageal pyomyositis in an intravenous drug user

Frank Cheau-Feng Lin, Kee-Ching Jeng, Stella Chin-Shaw Tsai
2014 Interactive Cardiovascular and Thoracic Surgery  
An inflammatory or infectious disease of the oesophagus occurring in tissue layers beneath but sparing the mucosa may pose a diagnostic challenge. Bacterial pyomyositis has been previously reported occurring mostly in the skeletal muscle. Pyomyositis involving the gastrointestinal tract is extremely rare, and may easily be misdiagnosed due to its nonspecific clinical features. We report a case of an intravenous drug user who presented with oesophageal pyomyositis. Early computed tomography
more » ... itated accurate diagnosis. Adequate drainage followed by antibiotic treatment was effective and the oesophagus was preserved. To the best of our knowledge, this is the first report of a case of oesophageal myositis in an intravenous drug user.
doi:10.1093/icvts/ivu269 pmid:25125141 fatcat:rgew3554mzhlnhq5vdhvokmy4e

Meibography Phenotyping and Classification From Unsupervised Discriminative Feature Learning

Chun-Hsiao Yeh, Stella X. Yu, Meng C. Lin
2021 Translational Vision Science & Technology  
The purpose of this study was to develop an unsupervised feature learning approach that automatically measures Meibomian gland (MG) atrophy severity from meibography images and discovers subtle relationships between meibography images according to visual similarity. One of the latest unsupervised learning approaches is to apply feature learning based on nonparametric instance discrimination (NPID), a convolutional neural network (CNN) backbone model trained to encode meibography images into
more » ... dimensional feature vectors. The network aims to learn a similarity metric across all instances (e.g. meibography images) and groups visually similar instances together. A total of 706 meibography images with corresponding meiboscores were collected and annotated for the use of network learning and performance evaluation. Four hundred ninety-seven meibography images were used for network learning and tuning, whereas the remaining 209 images were used for network model evaluations. The proposed nonparametric instance discrimination approach achieved 80.9% meiboscore grading accuracy on average, outperforming the clinical team by 25.9%. Additionally, a 3D feature visualization and agglomerative hierarchical clustering algorithms were used to discover the relationship between meibography images. The proposed NPID approach automatically analyses MG atrophy severity from meibography images without prior image annotations, and categorizes the gland characteristics through hierarchical clustering. This method provides quantitative information on the MG atrophy severity based on the analysis of phenotypes. The study presents a Meibomian gland atrophy evaluation method for meibography images based on unsupervised learning. This method may be used to aid diagnosis and management of Meibomian gland dysfunction without prior image annotations, which require time and resources.
doi:10.1167/tvst.10.2.4 pmid:34003889 pmcid:PMC7873493 fatcat:7rc6mabqyfhxfgg2m7d4j4t7va
« Previous Showing results 1 — 15 out of 6,237 results