Linguistic knowledge is of great benefit to scene text recognition. However, how to effectively model linguistic rules in end-to-end deep networks remains a research challenge. In this paper, we argue that the limited capacity of language models comes from: 1) implicitly language modeling; 2) unidirectional feature representation; and 3) language model with noise input. Correspondingly, we propose an autonomous, bidirectional and iterative ABINet for scene text recognition. Firstly, the autonomous suggests to block gradient flow between vision and language models to enforce explicitly language modeling. Secondly, a novel bidirectional cloze network (BCN) as the language model is proposed based on bidirectional feature representation. Thirdly, we propose an execution manner of iterative correction for language model which can effectively alleviate the impact of noise input. Additionally, based on the ensemble of iterative predictions, we propose a self-training method which can learn from unlabeled images effectively. Extensive experiments indicate that ABINet has superiority on lowquality images and achieves state-of-the-art results on several mainstream benchmarks. Besides, the ABINet trained with ensemble self-training shows promising improvement in realizing human-level recognition. Code is available at https://github.com/FangShancheng/ABINet.
Chemosensory proteins (CSPs) are a group of small soluble proteins found so far exclusively in arthropod species. These proteins act in chemical communication and perception. In this study, a gene encoding the Type 1 CSP (BtabCSP1) from the agricultural pest Bemisia tabaci (whitefly) was analyzed to understand sequence variation and expression specificity in different biotypes. Sequence analysis of BtabCSP1 showed significant differences between the two genetically characterized biotypes, B and Q. The B-biotype had a larger number of BtabCSP1 mutations than the Q-biotype. Similar to most other CSPs, BtabCSP1 was more expressed in the head than in the rest of the body. One-step RT-PCR and qPCR analysis on total messenger RNA showed that biotype-Q had higher BtabCSP1 expression levels than biotype-B. Females from a mixed field-population had high levels of BtabCSP1 expression. The interaction of BtabCSP1 with the insecticide thiamethoxam was investigated by analyzing the BtabCSP1 expression levels following exposure to the neonicotinoid, thiamethoxam, in a time/dose-response study. Insecticide exposure increased BtabCSP1 expression (up to tenfold) at 4 and 24 h following 50 or 100 g/ml treatments.
Aiming to represent user characteristics and personal interests, the task of user profiling is playing an increasingly important role for many real-world applications, e.g., e-commerce and social networks platforms. By exploiting the data like texts and user behaviors, most existing solutions address user profiling as a classification task, where each user is formulated as an individual data instance. Nevertheless, a user's profile is not only reflected from her/his affiliated data, but also can be inferred from other users, e.g., the users that have similar co-purchase behaviors in e-commerce, the friends in social networks, etc. In this paper, we approach user profiling in a semi-supervised manner, developing a generic solution based on heterogeneous graph learning. On the graph, nodes represent the entities of interest (e.g., users, items, attributes of items, etc.), and edges represent the interactions between entities. Our heterogeneous graph attention networks (HGAT) method learns the representation for each entity by accounting for the graph structure, and exploits the attention mechanism to discriminate the importance of each neighbor entity.
Through such a learning scheme, HGAT can leverage both unsupervised information and limited labels of users to build the predictor. Extensive experiments on a real-world e-commerce dataset verify the effectiveness and rationality of our HGAT for user profiling.
BackgroundUp to now, numerous case-control studies have reported the associations between fat mass and obesity associated (FTO) gene rs9939609 A/T polymorphism and polycystic ovary syndrome (PCOS), however, without a consistent result. Hence we performed current systematic review and meta-analysis to clarify the controversial results.MethodsCase-control studies reporting the relationship of rs9939609 A/T polymorphism and PCOS published before April 2015 were searched in Pubmed database without language restriction. Data was analyzed by Review Manager 5.2.ResultsA total of five studies involving 5010 PCOS patients and 5300 controls were included for further meta-analysis. The results of meta-analysis showed that the FTO gene rs9939609 A/T polymorphism was significantly different between PCOS group and control group in different gene models (For AA + AT vs. TT: OR = 1.41, 95% CI = 1.28–1.55, P < 0.00001. For AA vs. AT + TT: OR = 1.54, 95% CI = 1.25–1.89, P < 0.0001. For AA vs. TT: OR = 1.74, 95% CI = 1.38–2.18, P < 0.00001. For A vs. T: OR = 1.36, 95% CI = 1.25–1.47, P < 0.00001, respectively) suggesting that A allele was a risk factor for PCOS susceptibility. Furthermore, subgroup analysis in Asian and Caucasian ethnicities also found significant association between rs9939609 A/T polymorphism and PCOS (In Asian subgroup: OR = 1.43, 95% CI = 1.29–1.59, P < 0.0001. In Caucasian subgroup: OR = 1.33, 95% CI = 1.08–1.64, P = 0.008)ConclusionThis meta-analysis suggests that rs9939609 A/T polymorphism of FTO gene is associated with PCOS risk, and that A allele is a risk factor for PCOS susceptibility simultaneously.
We report a multiple laser stealth dicing (multi-LSD) method to improve the light extraction efficiency (LEE) of InGaN-based light-emitting diodes (LEDs) using a picosecond (Ps) laser. Compared with conventional LEDs scribed by a nanosecond (Ns) laser and single stealth-diced LEDs, the light output power (LOP) of the LEDs using multi-LSD method can be improved by 26.5% and 11.2%, respectively. The enhanced LOP is due to the increased side emission from the large-area roughened sidewalls of the sapphire substrates fabricated in the multi-LSD process. Numerical simulation results show that the multi-LSD process has little thermal damages to the multiple quantum wells (MQWs) of the LEDs.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.